ED 105 099

AUTHORTITLE

INSTITUTION

SiJNS AGENCY

REPORT NOPUB DATENOTEAVAILABLE FROM

EDRS PRICEDESCRIPTORS

DOCUMENT RESUME

CE 003 348

Bluestone, Barry A.The Personal Earnings Distribution: Individual andInstitutional Determinants. Final Report for thePeriod July 1, 1970--October 31, 1973.Michigan Univ., Ann Arbor. Inst. of Labor andIndustrial Relations.Manpower Administration (DOL), Washington, D.C.Office of Research and Development.DLMA 91-24-70-51-1Sep 74432p.; Ph.D. Dissertation, University of MichiganNational Technical Information Service, Springfield,Virginia 22151 ($6.00)

HF-$0.76 HC-$22.21 PLUS POSTAGEDoctoral Theses; *Economic Research; Employees;Employment Problems; Human Resources; IndustrialPersonnel; Industry; Labor Force; Labor Market; LaborUnions; *Low Income; Manpower Utilization; MultipleRegression Analysis; Occupational Mobility; *ResearchMethodology; *Salary Differentials; *Theories;Unskilled Workers; Wages

ABSTRACTThe study investigates the determinants of the

earnings distribution in the U.S. paying particular attention to theless-skilled segment of the workforce. A general earnings theory isproposed which has elements of human capital theory, institutionalhypotheses, and radical stratification analysis. Much attention ispaid to testing the "crowding" hypothesis that workers restricted toemployment in A limited number of industries or occupations will bepaid substantially less than workers who are not so restricted. Theregression results, based on a large integrated micro-macro data set,yield extensive evidence of stratification and industry variablesaffecting earnings after controlling for differences in humancapital. This is especially true among less-skilled workers. Theresearch includes both a literature review and extensivebibliography. (Author)

.f.

Report DLMA 91-24-70-51-1

THE PERSONAL EARNINGS DISTRIBUTION: INDIVIDUAL

AND INSTITUTIONAL DETERMINANTS

Barry A. Bluestone

Institute of Labor and Industrial RelationsUniversity of Michigan-Wayne State UniversityAnn Arbor, Michigan 48104

Septe.per 1974

U S DEPARTMENT OF SEALS'SEDUCATION & WELFARENA TIC- 'I- INSTITUTE OF

EDUCATIONDOCUMENT r+AS BEEN REPRO

OuCE 0 EX:.; I, Y CI' r.I O I ROM..E Pf RSON ON ORIGIN

,NG IT POINT :JI 4rt 4..,OR OPINIONSSi .5110 )O NOT MCI y RE PRE.

st-ht or FICos, NAT.,)NA. rNs*Iitif I 0(C)0C0I.ON Ps, not. OR POI +EY

Final Report for Period July 1, 1970 - October 31, 1973

Prepared for

U. S. Department of LaborManpower AdministrationOffice of Research and Development

Washington, D.C. 20210

LI) 2CJ

Report DLMA 91-24-70-51-1

THE PERSONAL EARNINGS DISTRIBUTION: INDIVIDUAL

AND INSTITUTIONAL DETERMINANTS

Barry A. Bluestone

Institute of Labor and Industrial Relations

University of Michigan-Wayne State University

Ann Arbor, Michigan 48104

September 1974

Final Report for Period July 1, 1970 - October 31, 1973

Prepared for

U. S. Department of Labor

Manpower AdministrationOffice of Research and Development

Washington, D.C. 20210

This report was prepared for the Manpower Administration,

U.S. Department of Labor, under research and development

grant no. DL 91-24-70-51. Since grantees conducting

research and development projects under Government sponsor-

ship are encouraged to express their own judgement freely,

this report does not necessarily represent the official

policy of the Department of Labor. The grantee is solely

responsible for the contents of this report.

t -! a

STANDARD TITLE PACE 1.

FOR TCC11N:CAL PrPORTS 9 I -?11-70-51- 1

3. !(....yo_ ( at slx No.

4. fide 3 Hcpo:: Dat..The Personal Earnings Distributioo: Individual Sep, 74.

and Institutional Detenni,lants 6. l'ericoseteg Organ...mh, cosh.

7. Aul.kot(.)

Rarry_Blo-stonc9. Petto:gaint, thtz.>niZat:ti Ad.!:

Jnstitute of Labor and 'Industrial RelationsUniversity of Michigan- Wayne State UniversityAnn Arbor, Michigan 48104

8. P:No:nu:à; 0:ga:4/atom Kept.No.

i0. Plow 1 o:k (hilt No.

11. Co:lac: ( :art No.

DL 9l -24 -70 -51

12. SpOrl.;A: :re 4.era, s N.U.S. DerarM,:ent of LnborManpower Ad:Anistra.,:ion

Office of'c'search and Dcvelolcment

1111 20th St., N.W. Washington, D.C. 20210

15. Supple me mar y

13. l)peofRepo:t&I-iod

Final

14. Spowsoling Agency Code

16. /.1.1:act,This study invcstivites the determinants of the earnings distribution

in the U,S, paying particular attention to the less-skilled segment ofthe workforce. A general earnincs theory is proposed which has elementshumaa capital theory, institutional hypothescs, cadical stratification

analysis. Much attention is paid to testing the "crowding" hypothesisthat workers restricted to employment in a limited number of industriesor occupations will be paid substantially less than workers who are not sorestricted, The regression results, based on a large integrated micro-macro data sct, yield extensive evidence of stratification and industryvariables affecting earnings i.fter controlling for differences in human

capital. This is especially true among less-skilled workers, The research

includes both a literature review and extensive bibliography.

17. Key lord`; Lio....raen: 1/u. I/est :sptor..c

Earnings; Employment; Labor; Nanpower; Unionization

17b. kentificfs/Open-Ende'd 'terms The Labor MarketLabor Force, Labor Market, and Labor RequirementsLabor Market Processes

17c- COCA Ti hiK:our 5C

18. Da:At:m:01 `tate:,ntDistribution is unlimited. Available from

Nation,.1 Teehrical Information Service, SpringfieldVa., 22151,

Faroe CFSTI-35 .47e>

19. :,ecuricy Class (Thl.Report)

17NCI-A"`ii'iLB26: :Net Of It (I hi

1).11).ITN( I Aqs11:11'1)

21. No. oi429pp.

22.

$6.001...:(.01mq DC ,50051 I

THE PERSONAL EARNINGS DISTRIBUTION: INDIVIDUAL

AND INSTITUTIONAL DETERMINANTS

by

Barry Alan Bluestone

A dissertation submitted in partial fulfillmentof the requirements for the degree of

Doctor of Philosophy inThe University of Michigan

1974

Doctoral Committee:

Professor Harold Levinson, ChairmanProfessor Daniel FusfeldAssistant Professor Malcolm CohenProfessor Gerald Gurin

ABSTRACT

THE PERSONAL EARNINGS DISTRIBUTION: INDIVIDUAL

AND INSTITUTIONAL DETERMINANTS

by

Barry Alan Bluestone

Chairman: Professor Harold Levinson

This study investigates the determinants of the earnings

distribution in the United States paying particular attention to the

less-skilled segment of the workforce.

A general earnings theory is proposed which has elements of

human capital theory, institutional hypotheses, and radical strati-

fication analysis. Much attention is paid t, testing the "crowding"

hypothesis that workers restricted to employment in a limited number

of industries or occupations will be paid substantially less than

workers who are not so restricted. It was hypothesized that after

controlling for differences in human capital, large wage differentials

would continue to exist for similarly qualified workers. These

differences could be attributed to the stratification of the labor

force, particularly by race and sex. Once stratified, differences in

industry characteristics would have an effect on the personal earnings

2

distribution as well. Those workers who gain employment in the

more concentrated, profitable, and unionized industries will earn

more than others who have similar work characteristics.

The regression results, based on a large integrated miclo-

macro data set, yield extensive evidence of stratification and industry

variables affecting earnings after controlling for differences in

human capital. This is especially true among the least skilled

workers in the labor force although there'is a substantial earnings

effect throughout most of the occupational hierarchy. While it was

impossible to obtain incontrovertible evidence that "crowding" was the

culprit in producing "human capital constant" wage differentials,

the evidence seems to point overwhelmingly in this direction.

Concentration and unionization also have a significant impact on

wages as well as a number of other industry factors.

The overriding policy implication following from this analysis

is that large scale government intervention is required in order to

correct the apparently massive "inefficiencies" that currently exist

in American labor markets. Intervention is required to equalize

human capital investment opportunities but equally important to break

down the barriers to inter-occupational and inter-industry mobility

that apparently still exist.

7

C) Barry Alan Bluestone 1974

All Rights Reserved

Reproduction by the United StatesGovernment in whole or in part ispermitted for any purpose.

The material in this project was prepared under

Grant No. 91-24-70-51 from the Manpower Administration,

U.S. Department of Labor; under the authority of Title I

of the Manpower Development and Training Act of 1962.

Researchers undertaking such projects under Government

sponsorship are encouraged to express freely their

professional judgment. Therefore, points of view or

opinions stated in this document do not necessarily

represent the official position or policy of the

Department of Labor.

iii

9

PREFACE

The present study began a number of years ago, when in the

course of poverty research, the common stereotype of the poor was

shattered by the revelation that the majority of the poor work

and in nearly a third of all poor households the head works full-time

all year round.1

'

2 The AFDC mother, the aged, the infirm, the

industrially "undisciplined," in short those out of the labor force

or unemployed were found to be only a portion of the poor. For many

others poverty was discovered to be the result of low-wage employment

rather than no employment at all.

Many of the particular causes which explain the poverty of the

nonworking poor--sickness, old age, illiteracy, and "bad luck"--fail

to adequately explain the poverty incomes of those who work. For

them poverty is a much more complex phenomenon going beyond individual

1Some of the earliest research on the working poor include:

George Delehanty and Robert Evdns, Jr., "Low-Wage Employment: An

Inventory and an Assessment" (Northwestern University, n.d.)

unpublished manuscript; Laurie D. Cummings, "The Employed Poor:

Their Characteristics and Occupations," Monthly Labor Review, July

1965; Dawn Wachtel, The Working Poor (Ann Arbor: Institute of Labor

and Industrial Relations, University of Michigan-Wayne State

University, 1967) mimeo; Barry Bluestone, "Low-Wage Industries

and the Working Poor," Poverty and Human Resources Abstracts, March

1968.

2Computed from "Work Experience of Family Heads, by Poverty

Status of Family, 1968," U.S. Department of Labor, Manpower Report

of the President (Washington: U.S. Government Printing Office),

March 1970, Table 1, p. 121.

iv

10

inadequacies. The confluence of market forces and personal

attributes forms a complex web from which the individual factors

contributing to low earnings are difficult to unravel. Wage theory

should help us understand the problem, but so far it has generally

failed.

The reason for this is.that the simplifying assumptions in

traditional wage theory tend to confuse the issue. The assumption of

a homogeneous labor force, found in the institutionalist framework,

tends to obscure the earnings effect of differences in skills and

competencies among workers. Alternatively assuming perfect competi

tion and labor mobility as in the pure human capital theory obscures

many other factors which impose their own order on the distribution

of income and earnings. An understanding of the working poor

requires a general wage theory that focuses on both the characteristics

of workers themselves, and on the labor markets in which they work,

while dropping the restrictive assumptions normally found in

traditional wage theory. To understand the determinants of low

wages requires an understanding of the whole distribution of earnings.

What began as a narrow study of poverty employment thus blossomed

into a more general investigation of the determinants of personal

earnings in the United States.

My original interest in the problem was spurred by Louis Ferman,

Director of the Research Division of the Institute of Labor and

Industrial Relations, University of Michigan-Wayne State University.

My colleagues Mary Stevenson and Charles Betsey helped prepare the

v

11

data set and contributed to some of the analysis. William Murphy

was indispensible in writing computer programs well beyond my

capability while Lynn Ware, James Sumrall, Jr., and Martha

MacDonald troubled over some of the theory and mathematical presenta-

tion with me. Countless friends associated with the Union for

Radical Political Economics were helpful at various times in

suggesting hypotheses to test and always kept steering my research

in relevant directions. Mrs. Kathleen Schwartz was responsible for

diligently typing the final draft. Finally a special note of

appreciation goes to my dissertation committee, Professors Harold

Levinson, Malcolm Cohen, Daniel Fusfeld, and Gerald Gurin. The

committee aided me immeasurably in the development of the project and

always did their best to force me to consider all sides of the issues

involved. To all of these people I extend my warmest appreciation for

their help and their friendship.

vi

12

CONTENTS

LIST OF TABLES

Page

.. ix

LIST OF ILLUSTRATIONS xii

Chapter

I. INTRODUCTION 1

Wage Theory and the Study of PovertyThe Design of the Study

II. EXISTING THEORIES OF WAGE DETERMINATION 11

Marginal Productivity TheoryHuman Capital TheoryThe Basic Human Capital ModelEmpirical Studies of the Human Capital

Earnings FunctionThe Institutional ApproachBalkanized Labor MarketsAn Institutional Model of Wage DeterminationEmpirical Verification of the Institutional

ApproachProblems with the Institutional ApproachSocial Stratification AnalysisSocial Stratificatioa and Wage Theory

Toward a Complete Theory of Wage Determination

III. A SOCIAL STRATIFICATION Tr*',,,RY OF PERSONAL

EARNINGS 64

The General Stratification ModelThe Specific ModelA Mathematical TreatmentThe Reduced FormThe Data Source and the Set of Regression Variables

IV. THE ESTIMATION PROCEDURE 111

Potential MulticollinearityThe Estimation Procedure MechanicsProblems in Parameter Specification

vii

13

CONTENTS (Continued)

Chapter Page

V. THE REGRESSION RESULTS 128

The Regression ResultsOccupation Stratum 1-3Occupation Stratum 5Occupation Stratum 6-9Occupation Stratum 12-14Occupation Stratum 15-17Cross Occupation StrataThe "Grand Pooled" Regressions

VI. AN EVALUATION OF THE REGRESSION RESULTS 223

The MethodologyThe Z* MeasureThe ResultsThe Relative Impact of Human Capital and

Non-human Capital Factors

VII. CONCLUSIONS AND IMPLICATIONS 274

Theories to Explain these ResultsWhat's To Be Done?

APPENDICES

A. THE DATA SOURCE 295

The SEO Data BaseConstruction of Occupation Scores

Occupation DataConstruction of Industry Data Base

Industry Data

B. MEANS, STANDARD DEVIATIONS, AITCORRELATION MATRICES 334

BIBLIOGRAPHY 405

viii

14

LIST OF TABLES

TablePage

4.1 Zero-Order Correlation Matrix- -White Males/

All Occupation Strata 114

5.1 Regression Equations: Occupation Stratum 1-3

by Race and Sex 135

5.2 Regression Coefficients on the Stratification

Variables in the Pooled OCC STRATUM 1-3 Regressions 146

5.3 Regression Equations: Occupation Stratum 5 by

Race and Sex 150

5.4 Partial (XtX) Matrix for Occupation Stratum 5 160

5.5 Regression Equations: Occupation Stratum 6-9 by

Race and Sex 164

5.6 Regression Equations: Occupation Stratum 12-14

by Race and Sex

5.7 Regression Equations: Occupation Stratum 15-17

by Race and Sex

174

186

5.8 Regression Equations: All Occupation Strata

by Race and Sex 191

5.9 Potential Wage Rates for Black Male Workers

Under Varying Assumptions about their

Characteristics 201

5.10 Potential Wage Rates for White Female Workers

Under Varying Assumptions about their

Characteristics 207

5.11 Potential Wage Rates for Black Female Workers

Under Varying Assumptions about their

Characteristics 210

!

ix

is

LIST OF TABLES (Continued)

Table Page

5.12a Estimated Hourly Earnings under VariousAssumptions Concerning the Human Capital

Module in the Occupation-Pooled Regressions:School Completed = 8 Years 213

5.12b Estimated Hourly Earnings under VariousAssumptions Concerning the Human CapitalModule in the Occupation-Pooled Regressions:School Completed = 12 Years 214

5.12c Estimated Hourly Earnings under VariousAssumptions Concerning the Human Capital

Module in the Occupation-Pooled Regressions:School Completed = 16 Years 215

5.13 Partial (XtX) Matrix for "Grand" PooledRegression 220

5.14 Summary of Significant %MININD Factors 221

6.1 Wage Intervals due to Stratification andIndustry Factors, By Occupation Stratum- -

White Males 238

6.2 Wage Intervals due to Stratification andIndustry Factors, By Occupation Stratum- -

Black Males 248

6.3 Wage Intervals due to Stratification andIndustry Factors, By Occupation Stratum- -

White Females 253

6.4 Wage Intervals due to Stratification andIndustry Factors, By Occupation Stratum-Black Females 257

6.5 Pooled Occupation Regression WageInterval Estimates 259

6.6 Wage Intervals due to Stratification endIndustry Factors, By Occupation Stratum- -

All Race-Sex Groups 260

6.7 Z/HC Ratios By Occupation Strata 267

x

16

LIST OF TABLES (Continued)

Table Page

A.1 The SEO Sample Population of FullTimeFull Year Workers 300

A.2 General Education Development 306

A.3 Specific Vocational Preparation Scale... 307

A.4 The Physical Demands Scale for Strength 307

AS "Occupation Levels" based on GED and

SVP Scores 310

A.6 Census Occupations by "Occupation Level" 312

A.7 Occupation Data by SEO Occupation Code 320

A.8 The Industry Data 331

I

xi

17

LIST OF ILLUSTRATIONS

Figure

2.1 Marginal Revenue Products with Labor Supply

Page

Restrictions 20

2.2 The Institutionalized Wage Bargain 41

3.1 A Social Stratification Model of the Personal

Earnings Distribution 71

3.2 No "Crowding" 74

3.3 Simple "Crowding" 75

3.4 Complex "Crowding" 76

4.1 The Theoretical Relationship betweenEarnings and Experience 123

xii

18

CHAPTER I

INTRODUCTION

What factors determine an individual's wage? How are earnings

related to "skills" and to "productivity." To what extent does the

distribution of personal earnings reflect the distribution of skills

and to what extent institutional factors? These are the fundamental

questions which are the concern of this dissertation.

There exists today no generally agreed upon wage theory.

Rather there exists a set of hypotheses, each constrained by its own

set of assumptions, each with its own distinct set of "exogeneous

variables," and each in competition with the other. Consequently,

there is general confusion over the relationship between "wage,"

"skill," and "productivity." Adam Smith's theory of "compensatory"

wage differentials, J.B. Clark's marginal productivity theory, and

the investment theory of earnings stemming from the work of Denison,

Schultz, and Becker all posit a tight relationship between an

individual's own skills, productivity, and earnings. In opposition,

institutionalist wage theory discards the neoclassical assumptions

of perfect product and labor markets thereby disrupting a more

perfect mapping of human capital into the distribution of earnings.

Industrial characteristics replace human capital attributes as the

primary variables in institutional wage theory. "Radical" wage

1

19

2

theory, based on social stratificationanalysis, goes beyond the

institutionalist critique, severing any link between personal human

capital investment decisions and individual earnings. An individual's

stock of human capital, according to radical theory, is a function

of class, race, and sex. Relative earnings are determined by social

status rather than individual choice in the acquisition of human

capital.

The competition between neoclassical, institutionalist, and

radical theories remains largely unresolved.1 Each theory has a

distinct wage generating function which explains only a portion of

the existing variance in earnings. Generally the theories have not

been tested against each other. Consequently, a wage theory synthesis

has not evolved, much less a new scientific "paradigm," to use the

terminology of Thomas Kuhn.2

Yet the framework for a synthesis can be constructed. By

substituting the assumptions of institutionalist and radical theory

into the overly restrictive model of the neoclassical paradigm, a

"flexible" general wage theory can be developed. Specifically

accounting for imperfect product and labor markets produces a wage

theory capable of defining the complex links between human capital,

industry and occupational attachment, and the distribution of earnings.

1For the best discussion of the competition, see David M.

Gordon, Theories of Poverty and Underemployment (Lexington: D.C. Heath,

1972).

2Thomas Kuhn, The Structure of Scientific Revolutions (Chicago:

University of Chicago Press, 1962).

20

3

In the general theory developed here a human capital earnings

function is modified to allow for institutional barriers to

industrial and occupational access along racial and sexual lines.

The independent effects of industry characteristics such as unioni-

zation, concentration, profits, and capital-intensity also enter into

the wage model.

Wage Theory and the Study of Poverty

A correct specification of the wage determination process is of

more than academic interest. A good ?art of the government's

antipoverty strategy of the 1960s was based on labor market studies.

Translated into public policy, human capital research contributed to

the emphasis on manpower and human resource development programs.

Along this line, the so-called war on poverty was designed to "find

new means for offering disadvantaged groups in urban and rural

America a chance to develop their own capacities and become productive

members of our society.0 Federal outlays for all manpower

activities rose steadily during the latter part of the decade in

response to the presumption "that education and training are

specially effective ways to bring people out of poverty."4 Programs

totalling only $184 million dollars in 1964 grew to nearly $2.4 billion

3,'The Budget Message of t

Printing Office, 1969), p. 47.

he President," The Budget of

States Government, Fiscal Year

4Thomas I. Ribich,

Institution, 1968), p. 1.

1970 (Washington, D.C.: U.S.

Educationion and Poverty (Washington:

21

the United

Government

Brookings

4

by 1970.5

The results of this effort were mixed. The government's

attempt to "upgrade," "rehabilitate," "train," "retrain," "integrate,"

"reintegrate," and "prepare" the poor for the job market was often in

vain. The payoff in terms of employment gains and increasing earnings

frequently failed to live up to expectations. No matter how measured,

the cost of a particular prpgram often exceeded the benefits. Many

programs had low job retention rates and often entrants did not

complete their training cycle. In other cases trainees completed a

manpower program only to find it impossible to gain adequate employ-

6ment.

Particular manpower programs failed because of insufficient

funding, lack of coordination, inadequate training, and poor

forecasting of job opportunities. But even the successes brought

little reason for enthusiasm. For those who completed MDTA training

in the middle of the 1960s, only three out of five advanced in pay,

and the increased earnings were quite small. According to the largest

study of MDTA, involving over 100,000 institutional training graduates,

5Sar A. Levitan, "Manpower Programs Under Republican

Management," Poverty and Human Resources, March-April, 1970, p. 12.

6In a comprehensive analysis of the institutional portion of

the MDTA program it was found that over 30 percent of the trainees

dropped out before completing vocational training and only 58 percent

found jobs related to their training curriculum. For a comprehensive

overview of the manpower programs in the 1960s, see Sar A. Levitan

and Garth L. Mangum, Federal Training and Work Programs in the Sixties

(Ann Arbor: Institute of Labor and Industrial Relations, 1969).

22

5

the average wage for males after training was $2.06 an hour,

27 percent higher than the average pretraining wage. For females,

the post-training wage was raised to $1.53, less than 20 percent

above pretraining levels and below the then prevailing federal minimum

wage.7 What is worse, these statistics apply only to those who

actually completed a manpower program and found jobs. Thousands of

other failed to complete programs and others finished and could find

no suitable employment.8

What explains the apparently low returns on investment in

manpower programs? One explanation, of course, is that existing

programs actually add little to the "human capital" of the disadvantaged.

Much more extensive human resource programs are necessary before

satisfactory returns can be anticipated. The other explanation rests

on the hypothesis that a lack of human capital is not the major barrier

to economic success for the poor. Augmenting an individual's stock

7U.S. Department of Labor, Manpower Administration, "The

Influence of MDTA on Earnings," Manpower Evaluation Report, No. 8

(Washington: U.S. Government Printing Office, December 1968), p. 18.

8In place of institutional manpower programs, on-the-job training

funded by the federal government has provided more people directly

with jobs. But according to a GAO report, the federally-funded on-the-

job training program is no more than a transfer system whereby the

government pays for specific job training which would normally be pro-

vided by the cooperating firm in spite of the program. The General

Accounting Office uncovered the fact that: "OJT contracts had served

primarily to reimburse employers for OJT which they would have con-

ducted even without the government's financial assistance. These

contracts were awarded even though the intent of the program was to

induce new or additional training efforts beyond those usually carried

out." See U.S. General Accounting Office, "Improvements Needed in

Contracted for On-the-Job Training under the Manpower Development and

Training Act" (Washington: U.S. Government Printing Office, 1968).

23

6

of human capital, it may be argued, yields an insignificant marginal

return because employment opportunities are nonexistent or highly

restricted. The "low-wage" workforce may possess the human capital

characteristics of higher paid members of the labor force, but fail

to earn larger incomes because of geographical immobility, the high

cost of job information, or racial and sexual discrimination. Low

relative earnings can result from the "crowding" of a subset of the

workforce into a limited number of industries and occupations.9

Denied access to other economic sectors for which they are qualified

on the basis of human capital, members of the "low-wage" workforce may

be competing with each other for the limited supply of jobs in the

sectors open to them. In this case, the maintenance of an "oversupply"

of workers in the "low-wage" sector may be the primary reason for low

wages, not a lack of human capital. In addition, the industries to

which economic minorities are restricted may consist of marginal firms.

operating within an economic environment characterized by low capital-

labor ratios, strong product market competition, and weak unions.

For any given degree of "crowding," firms in less "permissive"

economic environments may offer lower wages. Poverty earnings will

then be a function of social "underemployment" rather than personal

"underinvestment."

"Relative Underemployment" can be said to exist when an individual

qualifies for higher wage employment on the basis of human capital but

9The "crowding" hypothesis can be traced to F.Y. Edgeworth,"Equal Pay to Men and Women," Economic Journal, December 1922.

L 24

7

is denied access to it on other grounds. If underemployment is

widespread among low-wage workers, the problem of low-wage employment

is then only partly the effect of inadequate human capital. In this

case manpower programs will have a limited ability to make improvements

in the earnings of the low-paid.

To what extent low wages reflect inadequate amounts of human

capital versus restricted access to employment opportunity can only

be ascertained through an empirical investigation which permits both

factors to simultaneously enter the analysis. This is the reason

for developing a testable "general" wage theory. Measuring the effect

of human capital on the wage rate relative to the effects of industry

variables and restricted employment opportunity is the necessary

prerequisite for understanding both the promise and the limitations

of manpower policy. Beyond this, the testing of a general wage theory

focuses attention on the factors which are most important in the wage

determination process for all members of the workforce. Such research

can empirically account for the major variables Thurow had in mind

when he concluded that "the distribution of human capital is an

important ingredient in the distribution of income, but it is not the

sole ingredient. The actual dispersion of income is much greater than

would be predicted by the distribution of human capital.u10

10Lester C. Thurow, Poverty and Discrimination (Washington:

The Brookings Institution, 1969), p. 109.

25

8

The Design of the Study

The dissertation proceeds in the following way to construct and

test a general model of wage determination. The major strands of a

general wage theory are discussed in Chapter II. Marginal productivity

theory, the institutional analysis, human capital theory, and social

stratification hypotheses are initially discussed. Each theory is

carefully weighed in order to glean material useful for developing a

testable wage determination model.

The general model is developed in Chapter III. A complete

theory of wage determination is first constructed which takes as its

premise a "deterministic" view of social relations. The distribution

of earnings is made a function of four exogenous variables: race,

sex, social class, and innate ability. Following this a specific

testable model is derived based on human capital, institutional, and

stratification hypotheses. The specific model is constructed so as

to hold human capital constant allowing the earning effect of industry

and occupation "crowding" to be measured. From this a reduced form

earnings generating function is developed. Chapter IV discusses the

econometric techniques used to measure the independent effects of

human capital and "crowding." Attention is paid to the potential

problem of multicollinearity and the statistical procedures used to

overcome them.

The statistical results follow in Chapter V. Regressions are

presented for five separate occupation groups which span the range of

all specific occupations in the United States. Individual regressions

i 26

9

are reported for each race-sex group as well as pooled race-sex

equations. In addition pooled regressions are presented which cover

the whole spectrum of occupations. The effect of individual human

capital, industry, stratification, and working conditions variables

is discussed.

An evaluation of the empirical results follows in Chapter VI.

Here the regressions are interpreted so as to parcel out the variance

in earnings due to human capital factors as a whole viz-a-viz labor

force stratification. Wage "ranges" are established for each

occupation group and each race-sex group based on a technique which

allows the human capital variables to be held constant while the

industry and stratification variables are permitted to vary together

according to empirically derived coefficients.

The final chapter is devoted to general conclusions and policy

implications. Emphasis is placed on the role of manpower policy in

the general antipoverty strategy. Some of the implications for

training programs and income maintenance schemes are explored.

Finally, there is some speculation as to the justification for the

present distribution of earned income, given the empirical results

found in this analysis.

There are two appendices in addition to the seven chapters.

Appendix A contains a description of the integrated macro-micro data

set with a discussion of its construction. Variables used in the

regression analysis are defined and the shortcomings in each is noted.

Appendix B contains the means and standard deviations for each

i Kwc.,44

10

regression as well as a complete set of zero-order correlation

matrices for all of the empirical results.

i 28

CHAPTER II

EXISTING THEORIES OF WAGE DETERMINATION

Individual prices reflect a near infinite set of past, present,

and even future events. Previous capital expenditures, the whole

galaxy of current prices of complementary and substitute products,

and expectations about future prices all impirge on the current

market value of each good. Consumer attitudes, changing tastes,

government subsidies, tariff policy, antitrust action and hundreds

of other factors interact to determine millions of prices.

Nevertheless, the key factors which determine the price of most final

and intermediate goods are well-known.

Yet economists have always been perplexed by one special case:

the price of labor.

Marshall, Pigou, Taussig, and other leading theorists weretroubled by the "peculiarities" of the labor market--the

fact that the worker sells himself with his services, that

his immediate financial n.,!ld may place him at a disadvantagein negotiating with employers, that he is influenced by

nonpecuniary motives, that he has limited knowledge of

alternative opportunities, and that there are numerous objective

barriers to free movement of labor.'

Numerous attempts have been made to fit the theory of wages

into a more general analysis of price. By assuming away a number of

1Lloyd Reynolds, The Structure of Labor Markets (New York:

Harper and Brothers, 1951), p. 2.

11

29

12

the "peculiarities" of the labor market, economists have treated

labor in the same manner as other productive inputs in the economy.

"The theory of the determination of wages in a free market is simply

a special case of the general theory of value," Hicks wrote. "Wages

are the price of labour; and thus, in the absence of control, they

are determined, like all prices, by supply and demand."2

The history of labor theory is rich in these abstract attempts,

but poor in empirical verification. The relative impact of supply

and demand forces on the wage rate remains, for the most part, a

mystery. Even Hicks admitted that "a long road has to be travelled"

before the concepts of wage theory "can be used in the explanation

of real events. H3 Barbara Wootton has responded that, "In practice

this road seems to have been not only long, but so exhausting that

few travellers have attempted it."4

Before setting out on this difficult road, it seems good

practice to review some of the theories developed in the past. Four

broad strands in the development of wage theory can be discerned:

(1) marginal productivity theory (2) institutional theory (3) human

capital theory, and (4) social stratification analysis. Most wage

theory fits into at least one of these categories.

2J.R. Hicks, The Theory of Wages (London: MacMillan, 1932), p. 1.

3Ibid., p. 10.

4Barbara Wootton, The Social Foundations of 1.7rage Policy (London:

Unwin University Books, 1955), p. 12.

30

13

What we are ultimately searching for is a theory that will

explain the empirically observed distribution of personal earnings.

Furthermore such a theory must be capable of describing and inter-

preting the relationship between the personal characteristics of the

individual worker and the wage he or she receives in the marketplace.

Of particular concern is the relationshipbetween the wages received

and a subset of personalcharacteristics which we shall call the

"endogenous productivitycharacteristics" of the individual.

By the term "endogenous productivitycharacteristics" we shall

refer to the innate and acquired physical and mental attributesof

the worker useful as inputs in the production process.This term is

synonymous with the term "human capital" when defined narrowly "as

an individual's productive skills, talents, and knowledge."5 This

new terminology is introduced because the term "human capital" has

been broadened in some recentliterature to include such factors as

race, sex, and the physicalattractiveness of the individual. While

these factors may be important in determining the distribution of

earnings, we find it useful to separate nut the personal character-

istics which would have no relationship to the distribution of earnings

in a "blind" economy--an economy in which theproductivity of an

individual was not related to color, sex, or physical beauty.6

Given

5Lester C. Thurow,Investment in Human Capital. (Belmont:

Wadsworth PublishingCo., 1970), p. 1.

6This, of course, should not be construed as to deny the

importance of these factors in the actual distribution of earnings.

31

14

this understanding the terms "endogenous productivity characteristics"

and "human capital" will be used interchangeably.

In the next section we shall introduce the term "endogenous

productivity" which relates to the potential output of an individual

given his or her endogenous productivity characteristics. Endogenous

productivity can be shown to be theoretically distinct from the more

common term, marginal productivity of labor.

Marginal Productivity Theory

Much of the debate over the distribution of earnings rests

on a more fundamental theoretical debate concerning the usefulness

of marginal productivity theory. Consequently the theory provides a

good starting point for any discussion of wage determination.

Orthodox or traditional wage theory rests on the fundamental

proposition that labor is paid its marginal product. "Workers are

paid according to how much they contribute to marginal increases in

output. If increasing the number of employed workers by one worker

would increase output by $5,000, workers should be paid $5,000."7

If

there are no additions to complementary inputs in the production

process, the wage of the "marginal" worker and all intramarginal workers

Physical attractiveness, for instance, may be the most important

personal attribute in some lines of work. Whether an individual

receives a particular job or not may be a function of physical

attractiveness and the actual market value of an individual in certain

occupations is a function of such factors as well. The distinction

between skill, for instance, and racial and sexual characteristics

is clear enough; physicalattractiveness falls into a grey area

somewhere in-between.

7Lester Thurow, Poverty and Discrimination, p. 26.

32

15

will be exactly equal to the full measure of any additional output.

This result follows according to marginal productivity theory

assuming no barriers to labor mobility, a homogeneous labor force,

and assuming that all employment yields homogeneous nonpecuniary

utility (or disutility). Where the product market is characterized

by reasonable competition and labor is freely mobile between employers,

the wage rate will equal the value of the marginal physical product.

wi= VMP1 = (MPP

i.11)

Where the product market is characterized by monopoly elements, the

equilibrium wage will equal the marginal revenue product.

wi= MRP = (MPP MR)

In either case an employer will not hire an additional worker if

revenue generated by that individual is less than the addition to his

wage bill. This assures the wage of labor will never be above its

marginal revenue product, at least as long as employers attempt to

maximize profits. The assumption of competition among employers for

labor services, on the other hand, assures the wage will not fall

below labor's marginal revenue product, and in the case of perfect

product markets, not below VMP.

Under conditions of monopsony in the labor market, labor is

paid less than its marginal revenue product.

wi< MRP

i

16

Monopsonistic employers face a rising supply curve rather than an

infinitely elastic supply of labor. Additional workers can only be

obtained by increasing the wage. In the absence of wage

"discrimination," the marginal cost of labor increases by more than

the wage bill paid to the marginal worker. The employer has to pay

the higher wage rate not only to the additional worker, but to all

of his workforce. Under these conditions, the marginal cost of labor

will lie above the wage rate, and equilibrium will therefore be

reached at a point where the marginal productivity of labor exceeds

the wage level.

The traditional analysis of wage determination has normally

been applied at either the level of the aggregate economy or at the

level of the firm; marginal productivity theory was not specifically

developed to explain the personal distribution of income. At the

level of economy, the supply of labor is assumed perfectly inelastic

or upward sloping. In this case the theory is useful as a theory of

aggregate wages. At the level of the individual firm, the supply of

labor is assumed perfectly elastic (with the exception of the

monopsony case) and the theory describes the level of employment in

each firm.

As long as there is perfect mobility of labor and labor is

homogeneous, there will exist a unique market clearing wage throughout

the economy. Each worker will receive exactly the same remunerationil

8This again assumes that all jobs have homogeneous nonpecuniary

utility. Where this assumption does not hold, "compensatory" wage

i '34

17

and each firm will hire just enough labor at this rate so as to keep

marginal revenue product (or the value of marginal product) from

falling below the market wage.

Classical marginal productivity theory paid little attention to

the characteristics of labor supply; labor was assumed homogeneous

throughout the economy or homogeneous within major categories or

broad occupations. The quality of labor was consequently accepted as

given. It was assumed that workers of a given quality could move

from one employer to another without interference. Relaxing the

labor homogeneity assumption, but retaining the assumption of perfect

mobility, yields a "modern" marginal productivity theory which is

theoretically capable of describing the personal distribution of

earnings among workers with different levels of human capital. As

Thurow has noted, "In an economy with perfect competition and in

equilibrium, the distribution of marginal products is identical with

the distribution of earned income."9

And,

the supply and demand for labor with differing skills

and knowledge would determine the marginal product of

each variety of labor. Individual earnings would equal

their marginal products, and the allocation of human

capital would determine the distribution of earnings.

differentials develop to account for differences in the nonpecuniary

advantages or disadvantages of particular jobs.

9Lester Thurow, Poverty.and Discrimination, p. 29.

10Ibid., p. 96.

35

18

Given free access to the labor and product markets consistent

with their human capital, workers will be distributed so as to

maximize the total value of output and in turn each worker's marginal

productivity and wage. If the equilibrium is disturbed in some way

(by the introduction of new technology, for instance), the labor force

will be reallocated so that once again the ordinal ranking of

endogenous productivity characteristics is consistent with maximized

output. From an efficiency standpoint, the wage structure under these

conditions will be optimal. From the viewpoint of "equity," the

personal distribution of earnings will be colinear with the distri

bution of human capital.

The colinear relationship between earnings and human capital

will occur whether the product market is characterized by perfect

competition or monopoly. However once monopsonistic elements are

introduced into the labor market, colinearity disappears. Labor of

given quality will receive less in the monopsonistic firm and the

differential will persist as long as mobility to other firms is

restricted. Where labor differs as to quality, the statistical

colinearity between human capital and earnings depends on which group

of workers is restricted to the monopsonized sector. In any case the

hypothesized link between endogenous productivity characteristics

and wage rates no longer holds.

The usefulness of the marginal productivity theory as a theory

of the distribution of earnings rests on the assumption of perfect

mobility of labor. To the extent that this assumption is violated in

1

frt0rib

10

19

the real world, the theory fails to adequately explain wage

determination. It fails for it specifies only one of the two critical

linkages between the distribution of earnings and the distribution of

human capital. The linkage between the wage rate and the marginal

revenue product of labor is well described by the theory. What is

not specified is the connection between MRP and the level of human

capital. This link relies on the nature of the labor market. It is

possible that every worker is paid his marginal revenue product at

the same time that his marginal revenue product bears no relationship

to his level of human capital. Given imperfect mobility of labor,

it is possible that:

(1) wi =MRPi

(2) MRP1 # f(Human Capital)

In this case a knowledge of the distribution of human capital would

be insufficient to describe the distribution of wage income.

At this point it is helpful to introduce a new term in order to

differentiate between the actual marginal revenue product of each

worker and the hypothetical marginal revenue product each worker would

receive if there were no barriers to labor mobility and the economy

were in equilibrium. This hypothetical marginal revenue product

shall be referred to as the "endogenous revenue product." The

endogenous revenue product of individual i (ERPi) is the marginal

revenue product individual i, possessing endogenous productivity

characteristics, C., would receive if he were to compete freely in the

20

labor market with all other workers with characteristics C.3 .

To clarify the distinction assume that all labor is of

homogeneous quality and there are two firms operating with identical

labor demand curves. Under the condition of perfect labor mobility,

all workers will earn a wage, wi, equal to the economy-wide marginal

revenue product, MRP*.11

Now introduce an arbitrary barrier to labor

mobility which results in three-fourths of the total labor force

being restricted to Firm B. (See Figure 2.1)

FIRM B

MRP*----

FIRM A

Employment in B F.B E* EA E* Employment in A

Figure 2.1 Marginal revenue products with

labor supply restrictions

11This result holds even if the two firms face different product

demand curves. In this case the total labor force will be allocatedso that wi = MRP

i* in each firm; shifts in the level of employment in

each firm will assure this result.

21

Under these circumstances the wage received by workers in Firm A will

be equal to MRPA while the wage received by workers restricted to

Firm B will be equal to MRPB. In this case the colinearity between

endogenous productivity characteristics and actual marginal revenue

product is nonexistent. The endogenous revenue product of each worker

is equal to MRP* while the actual marginal revenue products are

MRPA and MRPB.

For workers in Firm A:

MRPw.=A

> MRP* E ERPi

For workers in Firm B:

w = MRPB

< MRP* E ERPi

To repeat, marginal productivity theory describes the link between

wiand MRP

i'but fails to describe the relationship between MRP

iand

ERPi. Thus the traditional theory cannot be used as an earnings

distribution theory where labor immobility is extensive. To summarize:

Under the assumption of perfect labor mobility:

(I) wi= MRP

(2) MRPi = ERPi

(3) ERPi= f(C )

(4) wi = f' (C )1.

Under the assumption of imperfect labor mobility:

(1) wi= MRP1

(2) MRPi # ERPi

22

(3) ERP1 = f(C )

(4) w #(C.i)i

)

and the marginal productivity theory is no longer a theory of the

distribution of earnings.

Human Capital Theory

Traditional marginal productivity theory rests on two funda-

mental assumptions: (1) homogeneous labor supply, and (2) perfect

labor mobility. Institutional wage theory, to be discussed in the

next section, retains the first assumption but rejects the latter.

Human capital theory does the reverse. It extends the marginal

productivity theory to account for differences in labor quality, but

maintains that all workers of a given quality compete in the same

market. In assuming no barriers to labor mobility (for labor of the

same quality), the human capital theory is fully consistent with

traditional wage theory. All workers who have the same endogenous

productivity characteristics produce the same marginal revenue product

and earn the same wage.12

Thus

= MRPij

= ERPwigij

for all individuals, i, with human capital characteristics, j. For

the labor force as a whole, the distribution of earnings becomes a

function of the distribution of human capital.

12This holds, once again, except in the case of monopsony.

40

23

Renewed interest in the role of labor in the production process

began in the late 1950s. Edward Denison, in attempting to account

for the sources of economic growth in the United States in the

1929-1957 period, found that he could explain only 31 percent of the

increase in output if he were forced to assume that productivity of

labor did not change.13

T.W. Schultz, in a classic article on the

same subject, showed that gains in output over time could not be

solely attributed to increases in physical capital.14

The need arose

for a framework which stressed "human productivity" as a source of

economic growth. The idea of human capital was introduced into

economic analysis.

The new approach to the study of wages and employment did more

than merely add the dimension of labor quality to the traditional

productivity theory. It focused on the investment process by which

a given stock of human capital is accumulated. "Human capital models,"

according to Mincer, "single out individual investment behavior as a

basic factor in the heterogeneity of labor incomes."15

Empirically,

the human capital approach attempts to measure the individual and

13Edward F. Denison, The Sources of Economic Growth in the United

States and the Alternative;, Before Us (New York: Committee forEconomic Development, 1962), p. 266.

14T W. Schultz, "Investment in Human Capital," American Economic

Review, March 1961. Also in "Investment in Man: An Economist's View,"

Social Service Review, June 1959.

15Jacob Mincer, "The Distribution of Labor Incomes: A Survey,"

Journal of Economic Literature, March 1970, p. 6.

41

24

social rates of return from investments in formal education, on-the-

job training, vocational education, health care, additions to labor

market information, and migration. Assuming away labor market

imperfections, some research is even attempting to infer the

distribution of abilities from the distribution of earnings.16

The Basic Human Capital Model

Borrowing from Mincer and Becker, .e fundamental human capital

equation can be written as in equation 1.

(1) log Ex

= log E0+ rx

where Ex= earnings generated by investment x

r = market discount rate or the internal rate of

return on investment x

E0= earnings generated by other factors than

investment x

According to this simple formulation the earnings distribution is a

function of investments in education, training, and other individually

acquired human capital components (x) plus a function of E0which

include:: innate or natural ability and other factors. Given a market

determined r and assuming E0 constant, differences in earnings will be

directly related to differences in the amount of human capital

acquired by each individual.

16Mincer notes two articles in this genre: K. Bjerke, "Income

and Wage Distribution," Review of Income and Wealth, November 1970;

A.D. Roy, "The Distribution of Earnings and Individual Output,"

Economic Journal, September 1950.

42

25

In addition, according to Mincer, equation (1) can account for

imperfections in the product market, in the labor market, and in the

market for human capital investment funds. Unfortunately, however, it

does this in patchwork fashion. As Mincer notes

If the competitive assumptions are relaxed, internal rates

of return cannot be equated with the market rate of interest

and generally differ among individuals. Equation (1) can

remain serviceable, however, with r interpr4ted as a group

average internal rate of return on (investment x), while

individual differences in r and in log E0 are impounded in a

statistical residual.17 (emphasis added)

This solution to the market imperfection problem is useful as

long as the degree of imperfection is minor.18 But if a large part

of the variance in individual earnings is due to labor market

imperfections, the human capital model fails to specify the critical

variables and in fact may draw attention away from them. A similar

"error" term or shift coefficient could have been applied to the

marginal productivity model, but there too, the patchwork would have

been for the sake of "realisM" without yielding any analytic

18Two points are in order. In private correspondence,

Professor Harold Levinson has noted that even this formulation by

Mincer is not quite correct. The group average r may also be affected

by labor market imperfections. Restrictions in supply for a whole

occupation, for instance as in the case of the building trades or

the medical profession, would shift r itself rather than show up in

the residual. In this case, he argues, equation (1) would be assigning

some portion of Ex to investment x which is in fact related to

institutional factors rather than inve--ment.

The second point regards the relationship of this formulation

to the differentiation between "endogenous revenue product" and

"marginal revenue product" discussed in the section on the marginal

productivity theory. Clearly ERP is analogous'to Mincer's "group

average internal rate of return" while the "statistical residual"

43

26

improvement in the model.19

Reinterpreting equation (1) indicates that there are two factors

of importance in wage determination. One is the amount of investment

in human capital and the other is the rate of return earned on the

investment. Or as Thurow has formulated:

The value of human capital can be divided into price and

quantity components. Education and on-the-job experience

provide the principal means for increasing the quantity or

quality of an individual's capital. Migration, improvements

in information, and the elimination of market imperfections,

such as prejudice, are the chief instruments to raise the

price for existing capital. Although the price factor

would not exist in perfect markets where all were paid equal

amounts for the use of identical skills, in imperfect

markets it is an important element in valuing human

capita1.2°

The distinction is important and lies at the crux of much confusion

over the usefulness of the human capital needed. Depending on the

degree of imperfection in the labor markets, the effect of r on Ex

accounts for variance in MRP around ERP. In this case, r = ERP while

(MRP-ERP) = c, thy. statistical residual.

191n the recent literature there has been some attempt at

explicitly integrating market imperfections into human capital theory.

Much of this has focused on job search behavior. The job search is

viewed as another form of investment in human capital where the costs

of the search, including opportunity costs, must be weighed against

potential discounted future earnings. While this tends to account

for the problem of "imperfect" markets due to information cost, it

fails to solve the larger problem of imperfect mobility due to market

discrimination. See Charles C. Holt and Martin H. David, "The Concept

of Job Vacancies in a Dynamic Theory of the Labor Market" (Madison:

Social Systems Research Institute, University of Wisconsin, 1965) and

Dale T. Mortensen, "A Theory of Wage and Employment Dynamics"; and

Donald A. Nichols, "Market Clearing for Heterogeneous Capital Goods,"

in Edmund S. Phelps, Microeconomic Foundations of Employment and

Inflation Theory (New York: Norton, 1970).

20Lester C. Thurow, Poverty and Discrimination, op. cit., p. 69.

44

27

may outweigh the effect of 1. How much each individual invests in

human capital may not be as important as the rate of return he or

she receives on that investment. For a given population the distri-

bution of r's may be such as to reduce significantly the covariance

between the distribution of x's and Ex's. In this case,

The concept of human capital loses its economic meaning.It no longer reflects productive capacities, and it nolonger can be viewed in the same light as physical capital.In a fundamental sense the problems of determining individualincomes cease to be economic and become sociological orinstitutional.21 (emphasis added)

Empirical Studies of the Human CapitalEarnings Function

The development of the human capital function was followed by

a steady flow of empirical studies aimed at quantifying the deter-

minants of earnings. Many of the earlier studies attempted to measure

the private and social returns to education by estimating the

discounted present value of investment in formal schooling.22

Other

21 Lester Thurow, Investment in Human Capital, op. cit., p. 18.

22Some of the more important studies in this area include:

Gary Becker, Human Capital (New York: National Bureau of EconomicResearch, 1964); W. Lee Hansen, "Total and Private Rates of Return on

Investment in Schooling," Journal of Political Economy, April 1963;

Orley Ashenfelter and J.D. Mooney, "Some Evidence on the PrivateReturns to Graduate Education," Southern Economic Journal, January1969; M. Blaug, "The Private and Social Returns to Investment inEducation: Some Results for Great Britain," Journal of Human Resources,

Spring 1967: A.B. Carroll and L.A. Ihnen, "Costs and Returns for TwoYears of Postsecondary Technical Schooling," Journal of PoliticalEconomy, December 1967; W. Lee Hansen, Burton Weisbrod, and W.J.Scanlon, "Schooling and Earnings of Low Achievers," American Economic

Review, June 1970; E.F. Renshaw, "Estimating the Returns to Education,"

Review of Economics and Statistics, August 1960; E.A.G. Robinson and

45

28

studies more explicitly analyze the distribution of income and

earnings with human capital factors as the independent variables.

Using diverse earnings functions, a number of studies have

attributed a large part of the explained variance in incomes to

differences in education. Morgan, David, Cohen, and Brazer, using

multiple classification analysis on national survey data, found the

most important factor determining hourly earnings of household heads

is an education-age interaction term.23

The beta Loefficient on the

education-age term was .234, highest among the fourteen variables in

their analysis including sex, occupation, race, geographic mobility

and several general demographic factors.24

Using the 1/1000 1960

Census sample, Giora Hanoch finds a relatively high "marginal product"

for education, although education appears in his formulation to be

subject to diminishing returns.25 The result is similar to Weisbrod's

findings for private rates of return on different levels of schooling.

J.E. Vaizey (eds.), The Economics of Education (Londol: St. Martins,

1966); Gerald Rose, Differential Returns to Investments in Human

Capital in the Academic Labor Market, University of California

(Unpublished Ph.D. dissertation, 1969).

23James N. Morgan, Martin H. David, Wilbur J. Cohen, and Harvey

E. Brazer, Income and Welfare in the United States (New York:

McGraw-Hill, 1962).

24Ibid., p. 48.

25See Giora Hanoch, "Personal Earnings and Investment in Schooling,"

Ph.D. dissertation, University of Chicago, 1965; also, "An Economic

Analysis of Earnings and Schooling," Journal of Human Resources, Winter,

1967.

i 46

29

Lowell Gallaway,26

H.S. Houthakker,27

Elizabeth Waldman,28

and

Herman Miller29

have all attached a great significance to formal

education in explaining the distribution of earnings.

More recently, however, a good deal )f research has called

into question the great importance of formal education in earnings

functions. This is especially true of studies directed at explaining

the income differences of whites and nonwhites. Using a 77 cell

education-occupation matrix, Bluestone et al. find that a maximum of

30.3 percent of the income differential between full-time, full-year

employed white and black men can be explained by the quantity of

formal schooling.30 Two-thirds or more of the differential is due

to occupational discrimination (education statistically held constant),

discrimination in industrial attachment, and human capital factors

not included in formal schooling. Only 2.8 percent of the total

differential between full-time employed white men and white women can

26Lowell Gallaway, "The Negro and Poverty," Journal of Business,January 1967, pp. 27-35.

27H.S. Houthakker, "Education and Income," Review of Economics

and Statistics, February 1959.

28Elizabeth Waldman, "Educational Attainment of Workers,"

Monthly Labor Review, February 1969.

29Herman P. Miller, "Annual and Lifetime Income in Relation to

Education," American Economic Review, December 1960.

30The calculations were made from data obtained in the 1967Survey of Economic Opportunity. These specific calculations can beobtained from the authors. Similar results in a more disaggregated

model can be found in Barry Bluestone, Mary H. Stevenson, and WilliamM. Murphy, Low Wages and the Working Poor (Ann Arbor: Institute of

r 47

30

be explained by formal schooling; 14 percent of the differential

between black women and white men. Donald Katzner found a similar

result for the white/nonwhite earnings differential.31

Michelson,

criticizing earlier studies for failing to account correctly for the

interaction between education and occupation, finds the effect of

schooling on earnings to be even smaller.32

Using a larger matrix,

Michelson shows "that only 16 percent of the earnings differential

between whites and nonwhites would have been corrected by equal

schooling categories, employing current (1959) earnings per year of

school for each racial group."33

In response to the evidence that formal schooling explains only

a fraction of the earnings differential, especially among race-sex

groups, additional human capital variables have been added to the

earnings function. A catalog of these variables compiled by Hansen,

Weisbrod, and Scanlon includes: (1) physical condition, including

general state of health and specific disabilities; (2) mental

capability, reflecting inherited potential; (3) learning and

experience, determined not only by the quantity and quality of formal

Labor and Industrial Relations, Research Division, University of

Michigan-Wayne State University, 1973).

31Donald A. Katzner, Theory and Cost of Racial Discrimination,

Ph.D. dissertation, University of Pennsylvania, 1966.

32Stephan Michelson, Incomes of Racial Minorities (Washington:

The Brookings Institution, 1968) unpublished manuscript.

33Ibid., pp. 2-35.

I 48

31

education, but by specific job training and job experience; (4)

psychological characteristics, among them motivation and ability to

communicate and cooperate in work situations; and (5) family

environment, reflecting informal learning, socialization, and

n contacts."34

The study by Morgan, et al. attempted to explain hourly earnings

using proxies for many of these factors. While "supervisory

responsibility," "attitude toward hard work and need-achievement

score," "interviewers' assessment of ability to communicate," and

"physical condition" were statistically significant, each of these

factors explained only a minute fraction of the variance in wage

rates after controlling for other variables.35

Altogether their

fourteen variables including education and age, sex, occupation, race,

and geographical location (in addition to the preceding variables)

accounted for 34 percent of the variance in wage rates. Two-thirds

remained unexplained.

Other researchers have continued to add new variables to the

basic human capital model in an attempt to explain the variance in

earnings. Chief among these are the quality of education, work

experience, and on-the-job training. Johnson and Stafford used

educational expenditure per pupil as a proxy for "quality" and found

34W. Lee Hansen, Burton A. Weisbrod and William J. Scanlon,

"Determinants of Earnings: Does Schooling Really Count?" Discussion

Paper, Institute for Research on Poverty, University of Wisconsin,

August 1968 (Preliminary Draft).

35Morgan et al., Income and Weifare in the United States,

Chapter 5.

49

32

"that there are high but diminishing marginal returns to investment

in school quality.06 Thurow has used years of experience in the

labor force as a proxy for "experience" or on-the-job training.37 Rees

and Shultz find "seniority" with the present employer to be the most

significant variable in explaining the wages of workers in their

Chicago labor market study.38 The use of "age" in other studies is

intended to act partially as a proxy for experience. In each case,

age, experience, general on-the-job training, or seniority has been

found significant. Yet with few exceptions even the most complete

human capital equation seldom explains more than 35 percent of the

variance in income and usually the explanatory power of such models,

measured in terms of R2

, is much lower.

Of course, a relatively high R2 only indicates correlation; it

indicates nothing about the causal nature of the relationship or even

the direction of causation. This is especially important for human

capital functions. In the case of experience, for instance, the

36George E. Johnson and Frank P. Stafford, "Social Returns to

Quantity and Quality of Schooling," Department of Economics, University

of Michigan, 1970, unpublished manuscript, pp. 17-18. Other works on

school quality as an input in the human capital equation include:

Finis Welch, "Measurement of the Quality of Schooling," American

Economic Review, May 1966; and James Morgan and Ismail Sirageldin, "A

Note on the Returns to Quality of Schooling," Journal of Political

Economy, September-October 1968.

37Lester C. Thurow, "The Occupational Distribution and Returns

to Education and Experience for Whites and Negroes," Federal Programs

for the Development of Human Resources, Joint Economic Committee

(Washington, D.C., 1968), pp. 267-84.

38Albert Rees and George B. Shultz, Workers and Wages in an Urban

Labor Market (Chicago: University of Chicago Press, 1970), pp. 84-85.

50

33

problem is severe. There are a number of interpretations of the

relationship between experience and earnings, all of which are con-

sistent with the data, but which point to extremely different hypotheses

about the determinants of income. One is that experience adds

directly to a worker's endogenous revenue product and therefore is a

legitimate human capital variable. Another, however, views experience

or seniority as reflecting nothing more than institutionalized pay

increments based on length of service and set out in collective

bargaining agreements or offered by employers to maintain morale. In

this case, "experience" may be totally unrelated to changes in an

individual's human capital.

Of even greater damage to the human capital interpretation of

"experience" is the possibility that the causal link between earnings

and experience may actually be reversed. Higher wages may cause

longer experience. Workers in "high wage" industries may have lower

turnover rates and therefore longer on-the-job experience because there

is little room for improving earnings by moving to new employment.

Workers in "low wage" industries or firms may feel less attachment

to their present employer and search often for new jobs. In this case

seniority may be low, but possibly the result of low wages rather than

the reverse.39 In addition, the existence of "internal labor markets"

39Doeringer has shown that this occurs often in ghetto labor

markets. In a study of a Boston manpower program, he found that job

tenure was directly related to wage level after controlling for other

factors. See Peter B. Doeringer, "Manpower Programs for Ghetto Labor

Markets," Proceedings of the 21st. Annual Winter Meeting of the

Industrial Relations Research Association, pp. 257-267.

51

34

produces a situation where training and experience become a function

of being hired.40

Similar ?roblems of interpretation exist with other human

capital variables as well. The relationship between years of formal

schooling and level of human capital is by no means clear. Bowles,

for instance, has begun work on educational production functions to

determine what inputs from formal schooling contribute to "pro-

ductivity."41

Of special concern is whether schooling actually con-

tributes to the endogenous revenue product of the individual or whether

the empirically derived relationship between education and earnings

merely reflects the use of formal schooling as a "credential" in the

employment screening process. Other human capital variables suffer

the same problems of interpretation. General and specific job

training, IQ scores, health and disability measures, and factors

contributing to geographical mobility all appear to contribute in one

way or another to the "explanation" of earnings. But the precise

connection between independent and dependent variables remains fuzzy.

The more important problem with the human capital theory remains,

however, even if the problem of causal relationships is set aside.

Like its predecessor in productivity theory, human capital models fail

40For the best discussion of internal labor markets, see Peter B.

Doeringer and Michael J. Piore, Internal Labor Markets and ManpowerAnalysis (Lexington: D.C. Heath, 1971).

41 See Samuel Bowles, "Towards an Educational Production Function,"in W. Lee Hansen (ed.), Education, income and Human Capital (New York:National Bureau of Economic Research, 1970).

52

35

as a theory of personal earnings where labor markets are imperfect.

The use of "race" and "sex" variables as human capital components or

"quality" variables clearly improves the fit of so-called human

capital functions. But it should be equally clear that such variables

are quite distinct from what we have called "endogenous productivity

characteristics." Where racial and sexual discrimination exist in the

labor market, or where mobility barriers are established through

monopsonistic power or trade union practice, the distribution of wage

income and the distribution of endogenous revenue products need not

be covariant. In this case explicit attention must be paid to the

institutional factors in the economy which impinge on the distribution

of earnings. If mobility barriers are important in the economy, the

ad hoc addition of new "human capital" variables may boost R2a bit,

but will add little to a meaningful explanation of wage determination.

The Institutional Approach

Whereas the marginal productivity theory and human capital

theory for the most part ignore the existence of barriers to labor

mobility, the institutional approach to wage theory begins with the

basic position that market imperfections are sufficiently widespread

to cause wage rates to deviate significantly from their free market

equilibrium levels. Thus, according to institutional theory, an

individual's actual marginal revenue product can diverge significantly

from his endogenous revenue product.

C3

36

The institutional approach, developed in the late 1930s, came

in response to rising unionism, a growing awareness of monopoly

elements in the economy, and increased government intervention in the

marketplace. Rather than an immediate concern with the determinants

of individual earnings, institutional theory has attempted to untangle

the various factors which impinge on interindustry and interregional

wage distributions. Where the marginal productivity theory focused

on absolute wage levels, the institutionalists have been more con-

cerned -with relative wages (and changes in relative wages) for similar

work. Assuming the "quality" of labor homogeneous within a given

occupational range, the institutional approach investigates the impact

of such factors as unionization and the effect of "ability to pay" on

relative wages. Lacking the theoretical rigor of other approaches

to wage determination, the institutional approach compensates with

vigorous empirical investigation.

Balkanized Labor Markets

Adam Smith attempted to explain wage differentials by noting

"compensating" differences in job content. In contrast, J.S. Mill

argued as early as 1847 that wage differentials are due to the absence

of competition in the market for labor.42 Stressing the existence of

42J.S. Mill, The Principles of Political Economy with Some of

Their Applications to Social Philosophy in Vol. II, Collected Works

of John Stuart Mill (Toronto: University of Toronto Press, 1965).

According to Mitchell, the first reference to "noncompeting" groups

in labor is found in Mill's lecture notes. See Wesley C. Mitchell,

Types of Economic Theory, Vol. I (New York: Augustus M. Kelley, 1967),

p. 562n.

54

37

barriers to occupational entry, Mill pictured the labor market as

deeply fragmented with individual workers falling into specifically

defined markets. Little intermarket mobility could be expected between

"noncompeting" groups of labor. Barriers to mobility between

occupations, in Mil: iew, were due to the social class structure

of who we would today call human capital investment behavior.

The theme of barriers to labor mobility is implicit in the

institutional anal} 's. Pat in place of strict family occupational

lines characteristic of a preindustrial era, the modern labor market

is "bulkanizea" or segmented into many sub-markets by institutionally

developed rules, both formal and informal.43

Entrance into each sub-

market, and movement within its internal market channels are often

strictly defined. Tha degree of unfettered choice within the overall

labor market is consequently diminished. Those who gain access to

restricted markets presumably gain woges higher than they normally

would in the face of perfect competition.

Restrict,,n of employment in any one sub-narket or firm occurs

in one of two ways (or both). In markets controlled by strong

employee organizations (e.g. building trades, the medical profession)

the actual supply of labor may be restricted to some given level.

The intersection between the market demand curve and the "institu-

tionalized" st ply curve yields the sub-market wage. In other markets,-...

43See Clark Kerr, "The Bulkanization of Labor Markets," in

E. Wight Bakke, Labor Mobility and Economic Opportunity (Cambridge:

The New 1,!chnology Press and John Wiley, 1954).

55

supply is not explicitly regulated, but the wage level is. In this

case the amount of labor employed in the sub-market is a function of

sub - market demand and the institutionally set wage. In both cases it

is presumed that the resulting wage exceeds the wage that would exist

in perfect competition.

The existence of strong unions on the labor supply side is an

important element in the institutional framewot.c. Yet other factors

which contribute to the balkanization of labor markets--racial and

sexual discrimination, barriers to geographical mobility, "lock-in"

effects of seniority, civil service channels, etc.--contribute to the

institutionalist argument that wage rates may reflect other factors

beside the'endogenous productivity characteristics of labor.

Balkanized labor markets, however, are only part of the insti-

tutionalist approach to wage theory. On the demand side, institu-

tionalists argue that firms do not profit maximize, that the marginal

conditions needed to maximize profits are not and cannot be known with

accuracy, and that firms have other goals with respect to their

workforces beside maximizing output at minimum labor cost.44

Instead

of reflecting the lowest wage possible in every instance, firms with

an "ability to pay" will often offer higher wages that necessary to

secure the quantity of a given quality of labor it desires. In the

44For the best summary of the institutionalist attack on

marginal productivity theory and for one of the strongest rejoinders,

see the Richard Leste.-Fritz Machlup "debate." This appears in three

issues of the American Economic Review, March 1946, September 1946,

and March 1947.

56

39

words of a leading institutionalist, "The major factor [in wage

determination) is differences in companies' wage - paying ability, plus

in some cases the presence of a union which forces a company closer

to the limit of its ability to pay."45

Thiq stress on "ability-to-pay"

has led many empirical studies to focu, on those factors related to

the well-being of a firm or industry: ,centration, profits,

capital-intensity, and productive efficiency. Unionization takes on

the role of a political power variable in addition to its role in

restricting labor supply.

An Institutional Model of WageDetermination

Pulling together the separate strands of institutional wage

theory allows the development of a unified institutional theory of wage

determination.46 Labor supply is assumed homogeneous in quality but

45Lloyd Reynolds, The Structure of Labor Markets (New York:

Harper and Brothers, 1951), p. 189.

46Some of the more important research which went into the insti-

tutional synthesis included: John Dunlop, Wage Determination UnderTrade Unions (New York: John Wiley, 1944); Arthur M. Ross, Trade Union

Wage Policy (U. of California, 1948); Harold M. Levinson, Unionism,

Wage Trends, and Income Distribution z_ 1914-1947 (Ann Arbor: University

of Michigan Press, 1951); Lloyd G. Reynolds, The Structure of Labor

Markets (New York: Harper and Brothers, 1951); Sumner Slichter, "Notes

on the Structure of Wages," Review of Economics and Statistics,

February 1950; Lloyd G. Reynolds and Cynthia H. Taft, The Evolution of

Wage Structure (New Haven: Yale University Press, 1956); William Bowen,Wage Behavior in the Postwar Period (Princeton: Princeton University,

Industrial Relations Section, 1960); Harold M. Levinson, PostwarMovement of Prices and Wages in Manufacturing Industries, JointEconomics Committee, 86th Congress, 2nd Session, Study Paper No. 21,

1960; Albert Rees, "Union Wage Gains and Enterprise Monopoly," Essayson Industrial Relations Research (Ann Arbor: University of Michigan,

57

40

balkanized by factors other than those related to endogenous

productivity characteristics. Product markets are differentiated by

entry barriers, either accounted for by scale requirements or spatial

area limitations.47 And unions are responsible for Picher directly

limiting labor supply or using their bargaining power to raise wage

levels above the competitive norm.

Wages are then determined through a complex interaction of

economic constraints, political decisions which affect the strength of

unionism, and finally the relative bargaining power of labor and

management.48

INDUSTRY WAGE DIFFERENTIALS

ECONOMIC FACTORS4t----- ------>BARGAINING POWER FACTORS41

POLITICAL FACTORS

At any given pcint in time, the general level of physical productivity

and market demand conditions place an upper limit on the final wage

Institute of Labor and Industrial Relations, 1961); H. Gregg Lewis,

Unionism and Relative Wages in the United States (Chicago: University

of Chicago Press, 1963; Harold M. Levinson, "Unionism, Concentration,

and Wage Changes: Toward a Unified Theory," Industrial and Labor

Relations Review, January 1967; Harold M. Levinson, Determining Forces

in Collective Wage Bargaining (New York: John Wiley, 1968).

47The argument about spatial limitations of the physical area

of a labor market is developed in Harold M. Levinson, "Unionism,

Concentration, and Wage Changes: Toward A Unified Theory," op. cit.

48This model is most thoroughly discussed in Harold M. Levinson,

Determining Forces in Collective Wage Bargaining, op. cit.

58

41

bargain. The hypothetical competitive labor supply curve places

the lower limit on the -sage bargain. The final wage settlement will

then lie within this range and be determined by the relative

bargaining power of employer and employee representatives and/or the

nature of the preference function of management where union strength

is either weak or nonexistent.

The institutional synthesis can be depicted as in Figure 2.2.

In the short run, firms are faced by a wage range, WW - Wm, where W

is the reserve price of labor (of a given quality) in the absence of

any restriction of labor supply. At a wage below Wc, firms will find

no one willing to work. Above Wm, firms will cease all production

because WM> MRP at all levels of output. Through collective

Employment

Figure 2.2 The institutionalized wage bargain

59

. .

42

bargaining, the labor supply curve will be raised vertically so that

the height of Su above We ,:epends on the relative bargaining strength

of labor and management. Increases in the marginal physical product

of labor (through increases in the capital/labor ratio or techno-

logical advances) or increases in marginal revenue (due to changes in

market demand conditions) will shift the MRP curve up and the wage

band will expand to WC M

- W' The final wage W* thus depends on economic

variables (marginal physical product and marginal revenue) and

bargaining power factors. Behind the scenes, the government plays a

political role in modifying the relative strength of labor and

management.49

Even with homogeneous labor supply, wage rates can differ

between firms or industries depending on the relative height and slope

of the marginal revenue product curves and the relative strength of

labor and management. The institutional model consequently predicts

the following results:

(1) Individuals barred from protected industries will earn lower

wages than individuals in other industries even where endogenous

revenue products are equal.

49In the long run W* is indeterminate without knowledge of the

elasticity of substitution between capital and labor. Autonomous

increases in the price of labor may drive firms to raise the capital-

intensity of their production processes. In this case, the Su curve(s)

and demand curves will no longer be independent. The precise wage

outcome requires information about the elasticity of substitution

between capital and labor and the wage preference function of union

leadership.

1 60

43

(2) Wages will reflect industry characteristics as well as the

endogenous productivity characteristics of the workforce.

(3) In particular wages of similarly qualified individuals will

be higher for those who gain access to unionized industries and

industries where competitive pressures are minimized through monopoly

or spatial limitations to market entry.

Empirical Verification of the

Institutional Approach

With few .ceptions data on manufacturing industries have been

used to test the institutional predictions. The usual dependent

variable is change in average straight-time hourly earnings over

time.50 The critical independent variables have been unionization

and concentration while some attention has focused on profits, change

in sales, value-added, and capital-labor ratios.51

50Changes or increasesin average wage rates rather than

absolute wage levels have been used in institutional analyses in order

to circumvent the problems caused by differences in the "quality" of

the labor force in different industries. Presumably while labor

quality may vary from industry to industry, changes in the average

endogenous productivity of an industry's workforce come only slowly.

Thus there should be little relationship between changes in wage rates

and changes in the quality of the labor force. Any significant corre-

lation between industry factors and changes in wage rates should

consequently be free of hidden correl:'tion with human capital variables.

Unfortunately this does not completely solve the problem, however.

51 For early work using some of these industry factors see:

Sumner Slichter, op. cit.; John T. Dunlop, "Productivity and the Wage

Structure," in Income, Employment and Public Policy, Essays in Honor

of Alvin H. Hansen (New York: Norton, 1948); Joseph Garbarino, "A

Theory of Inter-Industry Wage Structure Variation," Quarterly Journal

of Economics, May 1950.

61

44

In an analysis of data for the 1923-40 period, Garbarino found

a zero-order correlation of .7702 between the rate of increase in

earnings and the degree to which output is concentrated in a few

firms.52 In a similar study of 50 industries covering the period

1933-1946 Ross and Goldner found a very strong correlation between

concentration and changes in average straight-time hourly earnings.53

Significant positive relationships were also found with unionization

and growth in employment, although Ross and Goldner admitted that they

could not disentangle the independent effects of concentration, employ-

ment growth, and unionization. They concluded that concentration and

growth provide a "permissive" economic environment within which unions

can appropriate a portion of monopoly profits. In a later multiple

regression analysis, Bowen confirmed the difficulty of untangling the

institutional variables, but found that wages rose more rapidly in

industries with rising employment, higher profits, higher concentration

ratios and stronger unionization.54 Finally Levinson confirmed the

importance of profits, concentration, and unionization.55 He found a

strong relationship between earnings, lagged profits, and 1954 concen-

tration ratios, but found no general relationship between union

52Joseph Garbarino, op. cit., p. 302.

53Arthur M. Ross and William Goldner, "Forces Affecting the

Interindustry Wage Structure," Quarterly Journal of Economics, May 1950.

54William Bowen, op. cit.

55Harold Levinson, Post-War Movement of Prices and Wages in

Manufacturing Industries, op. cit.

62

4c5

strength and wage changes. The importance of union strength per se,

he argued, does not show up statistically, but nevertheless exists

through pattern and demonstration effects.

H. Gregg Lewis has derived numerous estimates of the effect of

unionization in different sectors of the economy f different

historical periods.56 Most of these estimates use data compiled by

a number of his students. Using data developed by Sobotka, Lewis

estimated that in 1939 unionization increased wages of common laborers

in the building construction industry by approximately 5 percent.

Skilled craftsmen, however, members of much stronger unions, were able

to raise their relative wage 25 percent through organizing.57

In

bituminous coal, the effect of unionism ranged from a high of over

120 percent in the early 1920s to zero at the end of World War II.58

The effect of unionization on relative wages in other industries

included 10-18 percent in rubber tire manufacturing (1948), 7 percent

in wooden furniture manufacture (1950), 19 percent for barbers (1954),

and 6-10 percent for hotel employees (1948).

The weighted average effect of unionism on wages in the 12

industry studies Lewis reviewed is 18 percent. In cross-industry

56H. Gregg Lewis, op. cit.

57The construction industry estimates derived by Lewis are

based on Stephen Sobotka, "The Influence of Unions on Wages and Earnings

of Labor in the Construction Industry," Ph.D. dissertation, University

of Chicago, 195'7..

58Based on Rush V. Greenslade, "The Economic Effects of

Collective Bargaining in Bituminous Coal Mining," Ph.D. dissertation,

University of Chicago, 1952.

63

46

global studies of interindustry wage variation, Levinson,

Garbarino, and Goldner found an effect of similar magnitude while

Ross and Ross and Goldner found a somewhat smaller effect (although

biased by the 1945-46 period). Levinson's estimate was in the

neighborhood of 17 percent while Garbarino and Goldner were closer to

15.59

Problems with The Institutional Approach

In relaxing the assumption of labor mobility inherent in tra-

ditional wage theory, the institutional approach should be an important

addition to an understanding of wage determination. Unfortunately,

however, there are a number of problems in institutional analysis which

detract from its general usefulness.

One of these is the difficulty in specifying and measuring such

abstract factors as "ability-to-pay" and "bargaining power." The use

of profit rates and concentration clearly do not serve as strong

proxies for either of these and the proi-ortion of employees covered by

collective bargaining agreements--the normal measure of "unionization"- -

certainly leaves something to be desired as a measure of restricted

59These estimates were derived by Lewis by correcting the

earlier estimates in the original studies in order to make them con-

sistent. For the detail on these studies, see, Harold M. Levinson,"Unionism, Wage Trends, and Income Distribution, 1914-1917," MichiganBusiness Studies (Ann Arbor:Bureau of Business Research, GraduateSchool of Business, University of Michigan, 1951); Joseph W. Garbarino,op. cit.; William Goldner, "Labor Market Factors and Skill

Differentials in Wage Rates," Industrial Relations Research Association,Proceedings of the Tenth Annual Meeting, 1958, pp. 207-16.

i 64

47

labor supply or bargaining power. Consequently even if industry

characteristics seem able to account statistically for differences

in wage levels or changes in wage levels, it is difficult to relate

the results of the reduced form equation to a specific theory of wage

determination.

A second problem, at least for present purposes, is that insti-

tutional wage theory does not account for the stratification of certain

workers into certa n industries. It generally assumes imperfect

mobility without specifying the parameters of stratification. While

there is some theory as to why certain industries become concentrated

or unionized, there is practically no hypothesis as to which workers

will be employed in the more unionized industries and which in more

competitive sectors of the economy. This problem, of course, stems

from the fact that institutional theory never was intended as an

explicit theory of the personal earnings distribution.

There is an additional problem, however, which is potentially

more serious than either of these. It is possible that the insti-

tutional model is partially misspecified.60

The correlation between

earnings and institutional variables may be partly spurious hiding a

60The term "misspecified" in this context does not simply refer

to the fact that one or more variables in the model may be specifiedin the wrong mathematical form (e.g. linear rather than log normal).Rather misspecification refers to the possibility that there doesnot exist a real causal relationship between the endogenous variableand the several exogenous factors. In this case the significantcorrelation found between variables is spurious. A correctly

specified model would be one in which the causal relationship betweenvariables is theoretically sound.

65

48

correlation between earnings and endogenous productivity character-

istics. Misspecification may result because of the sequential

ordering of the acquisition of human capital, the determination of

occupation and industry, and the receipt of earnings. If occupational

and industrial attachment is a function of the level of human capital,

some of the variance in earnings normally attributed to industry

characteristics may in fact be due to differences in human capital

factors correlated with industry variables, but unspecified in insti-

tutional models. More formally, to the extent that (1) this corre-

lation exists and (2) the acquisition of human capital is causally

prior to job placement, significant coefficients on industry variables

are spurious and due to errors in the specification of the institutional

model. In the extreme case, the true relationship is between human

capital and earnings and the institutional model vanishes.

The use of data on wage ch;:ages rather than wage levels does not

completely eliminate this problem. The extent of unionization or

concentration across industries may be perfectly colinear with the

industry distribution of human capital. In this case both industry

and human capital variables would equally describe interindustry wage

changes and there would be no way a priori to determine which is the

true relationship. It may be true that larger wage increases as well

as higher wage levels are accorded higher skilled workers.

A fair test of the effect of institutional variables thus

requires a model which explicitly accounts for differences in human

capital and furthermore specifies the relationship between industry

49

factors and endogenous productivity characteristics. Such a model

would improve our understanding of interindustry wage differentials.

To go even further and develop a complete and coherent theory of the

distribution of personal earnings requires additional information

on the social stratification process in labor markets.

Social Stratification Analysis

Sociological literature since the time of Marx is replete with

attempts to understand the development and structure of social

stratification. Marx's theory of class conflict placed stratification

at the root of all social change.61

He viewed man's relation to the

means of production as the primary determinant of economic structure

and class differentiation. Emile Durkheim similarly placed great

emphasis on the division of labor.62

Increasing social density, he

argued, led to increasing occupational differentiation, lessened

social consensus, and altered the nature of social solidarity.

Other sociologists have studied the nature of status and prestige.

Weber investigated the relationship between social and economic

orders, stressing the importance of status as differentiated from

economic standing.63

Others have attempted to distinguish t,.t

61For an excellent review of Marx's theory of class differ-

entiation, see Reinhard Bendix and Seymour M. Lipset (eds.), Class,Status and Power, "Karl Marx's Theory of Classes" (Glencoe, FreePress, 1953), pp. 6-12.

6 2Emile Durkheim, On the Division of Labor in Society (New York:

Macmillan, 1933).

63Max Weber, "Class, Status, and Party," in Max Weber, Essays in

Sociology (New York: Oxford University Pref.s, 1946).

50

differences between class, status, and prestige. Within this broad

thrust is a more specific inquiry into the roots of social and

occupational mobility.64

For present purposes, a narrower perspective on social strati-

fication is ne:essary. A general theory of the personal earnings

distribution requires that the mechanisms of occupation, industry,

and wage stratification be clearly described. While this cannot be

done here in detail, a brief taxonomy of economic stratification is

helpful.

Two distinct mechanisms of differentiation can be identified.

One follows from human capital theory and the other from the insti-

tutional analysis. The former relates to differential access to human

capital; the latter to differences in the rates of return on a given

set of endogenous productivity characteristics. Both are related to

differences in race, sex, and social class.65

(1) Access 'to human capital. Investment in human capital can

vary between individuals for numerous reasons. Differences in time

preference, for one, can make a large difference in how much and when

64See, Pitirim A. Sorokin, Social Mobility (New York: Harper,

1927). For a review of social mobility studies, see S.M. Miller,"Comparative Social Mobility," Current Sociology, 1960.

65By the term "social class" we shall refer to a group of

individuals who generally possess common economic and social character-istics. The key determinants of social class by this definition includeincome, wealth, consumption, and social status. Social status isnormally conferred through one's occupation. The intergenerationaltransfer of "social class," for empirical purposes, is measured byoccupational standing and/or income.

68

51

individuals invest in themselves. Differences in ability may affect

investment rates as well. Individuals with greater innate ability,

for instance, may tend to remain in school longer and invest in more

training especially if training and ability interact to produce

extraordinary returns. Differences in income-leisure preference may

also tend to differentiate human capital investment. In each of these

cases, differential amounts of investment may be said to be

"voluntary."66

There are other cases, however, where differential investment is

involuntary and reflects social stratification. Because human capital

investment funds are not a "free" good and the market for investment

funds is imperfect, social class, race, and sex can enter into the

determination of each individual's stock and structure of acquired

human capital. The level of private investment often reflects the

level of personal income while the level of social investment (e.g.

thrimck public schools) is dependent on the social stratification

among legal jurisdictions.

Given imperfect human capital markets, wages can differ con-

siderably among individuals even if innate abilities and personal

preferences are identical, product and labor markets are perfect, and

66Extreme caution must be exercised in the use of the term"vcluntary." An individual's time preference, for instance, probablydepends on a whole set of factors, many of which he cannot control.Family attitudes, the social millieu, and economic conditions mayall p.ay a role in determining an individual's time preference andincome-leisure trade-off. "Culture" obviously influences a person'smotivation and may well be associated with social class.

69

52

rates of return are equivalent everywhere for identical levels of

human capital. Under these conditions, if acces. to human capital

is influenced by such characteristics as t-ce, sex, and social class,

wage differentials will reflect these factors.

(II) Differential rates of return. Differential amounts of

investment are sufficient to produce a stratified wage structure. Yet

there is considerable evidence that suggests that wage rates differ

extensively among individuals with similar endogenous productivity

characteristics. These wage differentials can be viewed as differences

in the rate of return on human capital investment.

Some of these differences may be related to traditional insti-

tutional factors such as unionization, concentration, spatial barriers

to market entry, and imperfect information. Beyond this, however,

lies the effect of social stratification on the structure of labor

supply. Rather than randomly distributed, rates of return seem to be

significantly related to race and sex. Discrimination in the labor

market can take a number of different forms each contributing in a

distinct way to differentiating rates of return.67

Wage rates (or rates of return) can differ among two individuals

who perform precisely the same job in the same firm. This type of

differential might be termed "pure wage discrimination." The more

complicated forms (and possibly the mole pervasive) involve restricted

67Thurow has attempted to catalog the several types of discrimi-

nation found in the labor market and analyze the effect of each. His

"catalog" can be found in Lester C. Thurow, Poverty and Discrimination,op. cit., p. 117.

70

53

access to particular industries, occupations, or firms. Occupational

and industrial stratification will result in different rates of return

for the same endogenous productivity characteristics and thus play a

potentially important role in the distribution of personal earnings.

How important restricted access is in determining the distribution

of rates of return can only be ascertained through empirical investi-

gation.

Social Stratification andWage Theory

Taken to its extreme,,social stratification theory stands in

direct contrast to neoclassical theory. Where the traditional analysis

focuses on maximization behavior subject to economic constraints,

stratification analysis ultimately places responsibility on the

"constraints" for determining economic outcomes. In terms of the

personal distribution of earnings, the key variables in social strati-

fication theory are beyond the control of the individual. To summarize,

they include:

(1) Opportunity for private investment in human capital(2) Opportunity for social investment in human capital(3) Restrictions to entry into specific occupations(4) Restrictions to entry into specific industries(5) Job discrimination within individual firms(6) Wage discrimination within individual jobs

By allowing these factors to enter the formulation of wage

theory, two things are accomplished. First, the barriers to mobility

stressed in the institutional analysis ("noncompeting" groups,

imperfect labor market in:ormation, unionization, and spatial

I 71

54

limitations to entry) are extended to include the obviously important

factor of discrimination in the labor market. Second, the inclusion

of the social stratification variables provides a framework for

understanding the specific distribution of jobs and earnings over the

distribution of persons. Traditional human capital theory fails to

adequately explain the distribution of endogenous productivity

characteristics while traditional institutional analyses fail to

specify which individuals will gain access to which sectors of the

economy. Social stratification theory thus may provide part of the

answer necessary to close the system used to explain personal earnings.

There is one important problem with stratification analysis,

however. Taken alone it provides no more than a description of the

wage distribution at a given point in time. In this sense it is not

a "theory" of wages per se. Its key parameters, race, sex, and social

class, are not particularly useful by themselves in analyzing changes

in the distribution of earnings. Consequently, stratification analysis

must become part of a more general theory of earnings if the theory

is to yield any more than a static description.

Toward a Complete Theory ofWage Determination

Each of the four existing theories of wages, at least in its

"pure" form, exhibits at least one critical shortcoming which prevents

it from fully explaining the observed distribution of earnings.

Marginal productivity theory fails to account for differences in

endogenous productivity characteristics and for barriers to labor

i 72

55

mobility. Human capital theory, while rectifying the problem posed

by a heterogeneous labor force, fails to pay adeq uate attention to

labor market imperfections. Institutional analysi

economic results of market imperfection, but fails t

control for differences in human capital or describe

personal characteristics responsible for differential

sub-markets. Finally, social stratification analysis,

focuses on the

o adequately

adequately the

access to labor

which neither.

assumes homogeneous labor supply or perfect labor market , is greatly

weakened by its inability to describe the dynamics of wage

determination.

Yet each theory provides a potentially vital element in the con-

struction of a general wage model. Productivity theory indicates thP

on average wages will bear a close relationship to labor's marginal

product. Further describing the supply side of labor markets, human

capital theory predicts that relative earnings will be related to

investment in endogenous productivity characteristics. Institutlo

theory poses the possibility that labor market imperfections will

impinge on the wage determination process in such a way that the

distribution of earnings (for individuals with similar human capital)

will partially reflect industry and occupational attachment. And

social stratification analysis extends the traditional institutional

analysis to account for variation in human capital investment and

different rates of return on capital due to differences in race, sex,

and social class.

nal

i 73

56

Each thus provides part of the catalogue of variables which

enter into the wage determination process. But the real problem is

determining how much of the wage distribution is best described by

each theory. The few empirical studies which have attempted to test

wage models using variables from more than one theoretical framework

have produced somewhat ambiguous results.

The most complete of these is Weiss's study of concentration

and labor earnings.68

Before controlling for personal characteristics

and other industry variables in his 1966 micro data study, Weiss

found annual earnings of male operatives in unregulated industries

to be significantly correlated with both "unionization" and concen-

tration.69 He reported that, "Unionization seems to raise annual

earnings by about 16 percent when concentration is low, but to have

no effect when CCR (concentration) is high. Concentration seems to

raise earnings by about 33 percent when unions are weak, but by only

13 percent when they are strong."70

After the addition of personal

characteristics (residence, race, age, education, family size, and

migration) and other industry variables (employment growth, size of

establishment, type of manufacturing, percent male employment, percent

68Leonard W. Weiss, "Concentration and Labor Earnings,"American Economic Review, March 1966.

69Weiss's actual regression result was:

Y = 1936 + 53.47 CCR + 23.74 U - .4426 UCCR R2= .0401

(280.5) (7.81) (4.16) (.1030) N = 5187Y = 4419

70Ibid., pp. 104-105.

i 74

57

skilled employees, percent employment nonwhite, and percent of

,'mployees residing in the South), the effects of unionization and

concentration decrease significantly. Unionization remaii,a

statistically significant (t = 5.0) but the coefficient falls to the

level where an industry that is 90 percent organized yields earnings

which are only 6-8 percent higher than an industry with only half of

the employees covered by collective bargaining agreements.

Concentration is now only barely significant (t = 2.1) and increased

annual earnings (resulting from a difference in CCR of 40 points-

20 vs. 60) amount to no more than 3 to 5 percent.71

After the

addition of the personal characteristics data, Weiss can explain about

34 percent of the variance in earnings. He concludes that, "The

effects of most industry characteristics are nonsignificant and often

of unexpected signs after personal characteristics are introduced.

In general, employers who for any reason pay high salaries receive

'superior' labor in the bargain. The general picture is one of fairly

efficiently working labor markets, even where substantial monopoly

may exist.H72

Weiss's results, however, hardly justify this optimistic con-

clusion. For one thing, Weiss specifies the "unionization" variable

at the industry level rather than at the micro level. Stafford has

shown that the average effect of union membership on relative wages is

71Ibid., p. 108.

72Ibid., p. 116.

i 75

58

10-16 percent after controlling for education, age, industry, city

size, region, and race when union membership is measured for the

individual rather than the industry.73 In addition, Johnson and

Youmans indicate that after controlling for age and education, the

effect of unicnism is actually double that found in early institu-

tional studies.74

In their study union membership increases relative

wage rates by 34.2 percent. One interpretation of this result is that

unionism is a substitute for more education. Unions insulate less

educated workers as well as younger workers from the usual effect of

education and age on wage levels. A correct specification of the

unionization variable might significantly change Weiss's results.

There are other weaknesses in the Weiss study which merit

attention. One weakness lies in his sample of industries. By

restricting his research to individuals employed in mining, con-

struction, manufacturing, transportation, communications, or public

utilities, he fails to account for the full variance in unionization

and concentration in the economy. Including workers in other services

and in wholesale and retail trade would have expanded the variance in

his industry variables and probably would have increased the signifi-

cance of these factors.75

73Frank P. Stafford, "Concentration and Labor Earnings: A

Comment," American Economic Review, March 1968.

74George E. Johnson and Kenwood C. Youmans, "Union Relative Wage

Effects by Age and Education," Industrial and Labor Relations Review,

January 1971.

75We would expect this result because it is generally known that

76

59

Even more important, the specification of the model leaves the

results ambiguous. Many of the personal characteristics in the model

have little or no relationship to endogenous productivity character-

istics and consequently muddy the interpretation that can be given to

"fairly efficiently working labor markets." The coefficient of

-681.30 on the dummy value for Negro (with t=7.3) indicates a signi-

ficant market imperfection due to some form of racial discrimination.

The same is true for the dummy value for Spanish surname.

Similarly there are a number of industry variables which are

significant and reflect market imperfections rather than differences

in the endogenous productivity characteristics in the labor force.

The sex composition of the workforce in an industry is significant

and in the extreme case of perfect sex segregation might yield a

difference in annual earnings of $619. Using a dummy variable for

durable manufacturing also makes little intuitive sense as a measure

of human capital although the coefficient is significant and accounts

for a $211 difference in annual earnings.

If these variables are considered as industry characteristics or

"stratification" factors rather than as human capital factors, the

degree of misallocation in the labor market is much greater than

Weiss 2stimates. Weiss adniits as much in his conclusion.76

the service industries and wholesale and retail trade industries arevery weakly unionized and relatively competitive at the same time

that they are generally low wage. These points should pivot the

regression line around giving a higher coefficient (higher slope) on

the industry variables.

76Weiss, op. cit., p. 115.

60

This does not necessarily imply that no misallocationresults from high-wage payments in concentrated industries.Labor "quality" in this study includes such personalcharacteristics as race, which may be quite irrelevant tothe objectively evaluated productivity of the laborer involved.It has been suggested that firms with monopoly power usepart of their profits to hire congenial or socially acceptableemployees, an option not Ivai_able to employers subject to morestringent competitive pressures. If so, the earnings of laborin monopolistic industries may still exceed its marginal-revenue product, even though they apparently approximate thevalue of its alternative product.

In this case the institutional factors explain a large part of the

variance in the personal earnings distribution after controlling for

endogenous productivity characteristics.

Two other recent studies appear to add some collaborative

evidence to this conclusion. In their study of the Chicago labor

market, Rees and Shultz controlled for age, education, experience, and

seniority.77

They found that among material handlers, the "mean wage

of nonwhites is twenty-nine cents below the mean wage of whites. . . .

The coefficient of the nonwhite dummy in the regression is a negative

thirty-one cents, indicating that only about two cents per hour

can be attributed to differences between nonwhites and whites in the

other characteristics that enter into the regression."78 The

addition of establishment variables to the Chicago labor market

regressions reduced the effect of the race dummy, but did not eliminate

its significance altogether. Such a result seems to indicate that part

77Albert Rees and George P. Shultz, Workers and Wages in an Urban

Labor Market (Chicago: University of Chicago Press, 1970).

78Ibid., p. 106.

78

61

of the variance in racial wage differentials is due to discrimination

in access to "high-wage" firms. The remaining differential arises

from wage discrimination within each firm.79

An even more recent study by Wachtel and Betsey, using multiple

classification analysis on micro data, found a large portion of the

residual variance in wage rates (after controlling for education,

experience, race, sex, age, and marital status) could be explained by

a composite "occupation-industry" variable.80

Region of employment,

city size, and union status were also significant after controlling

for personal characteristics.

While both of these studies find large wage differentials

related to labor market barriers, neither indicates precisely what

factors operate on the demand and supply sides to produce this result.

Rees and Shultz used dummy variables to account for "industry" while

Wachtel and Betsey relied on dummy variables for occupation - industry

combinations. Neither study addresses the question of what industry

factors--higher profits, concentration, restricted access, etc.- -

are responsible for the significant coefficients on "industry."

Beyond the specification problem there is an even greater

weakness in all of these recent attempts to generate wage functions:

none develop an explicit comprehensive wage theory with which to

79For similar results see Alice Kidder, "Interracial Comparisons

of Labor Market Behavior," Ph.D. dissertation, M.I.T., 1967.

80Howard M. Wachtel and Charles Betsey, "Employment at Low Wages,"

The Review of Economics and Statistics, May 1972.

79

62

interpret the reduced form results. Consequently it is difficult,

if not impossible, to disentangle the particular effects of human

capital variables, other personal characteristics, industry "ability-=to-pay" factors, industry and occupation access barriers, and "pure"

discrimination. To disentangle these factors and explore the

determinants of the personal earnings distribution requires the

development of an explicit model and a set of data which includes

specific variables for human capital factors, industry characteristics,

and stratification effects.

In essence our quest is to distinguish what forms of labor

market segmentation are responsible for the large wage differentials

we find between individual workers in the U.S. economy. Segmentation,

by our definition, is simply the division of the labor force into

"non-competing" groups for any reason: real human capital differences,

unequal access to occupations and industries, and differential wages

for precisely the same job are all bona fide forms. One particular

form of segmentation, however, is singled out for special attention.

This form is "stratification" and refers to segmentation based on

non-human capital factors. It can be said to exist whenever the labor

force is divided on the basis of race, sex, social class, or by insti-

tutional factors such as differential access to union membership.

Stratification, however, takes a number of forms itself, one of which

i.s "crowding" where workers have differential access to occupations

and industries while another is pure wage discrimination within the

same industry or occupation. Distinguishing between the effects of

1 80

63

"crowding" and "pure discrimination" is often difficult, but an

attempt at empirically isolating the two can be made.

In the following chapter, a general model of wage determination

is constructed which draws on the strengths of each theory and

indicates how the dynamics represented in each theory interact to

produce the observed wage structure. In Chapter V a part of this

theory is then tested. Attention will focus on the factors which

affect the variance in rates of return on given human capital

characteristics rather than on the process by which those character

istics are acquired. Thus the analysis will primarily attempt to

evaluate the effect of stratification on the personal earnings

distribution, given the existing distribution of human capital.

81

CHAPTER III

A GENERAL THEORY OF PERSONAL EARNINGS DISTRIBUTION

In this chapter a general theory of the personal earnings

distribution is developed. The theory is based on concepts of social

stratification, institutional economics, and human capital (which in

turn embodies the chief tenets of marginal productivity theory).

For purposes of empirical testing, a specific wage function based

on the "crowding" hypothesis is developed and a reduced form is

generated that is consistent with the overall theory. This is done

in order to test the social stratification and institutional elements

while holding constant the human capital factors.

The neoclassical formulation of economic problems normally

assumes: (1) rational individual decision-making, and (2) utility or

profit maximization subject to constraint. Both are inherent in the

marginal productivity theory and form the foundation for the human

capital approach to wage determination.) Accordingly, individuals

)In his survey article, Mincer is particularly clear on this

matter. He notes that ". . . an important attraction of this theory

is that it relies fundamentally on maximizing behavior, the basic

assumption of general economic theory." Jacob Mincer, "The

Distrib ution of Labor Incomes: A Surve:,," Journal of Economic

Literature, March 1970, p. 23. David Gordon has summarized this point

as well.

"In emphasis if not in precise substantive hypothesis, thetheories seem to suggest that individuals have a nearly

64

82

65

make two critical decisions which determine their own income. Each

worker decides how much to invest in the accumulation of human capital

stock and how much time to devote to work and how much, to leisure.

In the investment decision, individuals are constrained by innate

ability, a diminishing marginal return to increments in investment,

and an inelastic supply of investment funds. In the labor-leisure

choice, the ultimate constraint is the number of hours in the day and

the physiological need for rest. Within a broad range, individuals

determine their own wage rate, the hours of labor they supply, and

conseyently their own annual and lifetime earnings.

In the general earnings distribution theory developed here, the

neoclassical assumptions are abaadoncd. To the extent that the tra-

ditional formulation rests on the concept of free choice or "free

will," the present model end-races the opposite philosophical position;

it is at root socially deterministic. The observed distribution of

earrings is not the product of numerous personal decisions, but rather

primarily a function of social class, race, an., sex. Ultimately,

these are the exogenous determinants of wage rates. In a social

unlimited range of opportunities in the course of their

lifetimes. This implication seems to play the same role in

theories of income as the notion of "consumer sovereignty"

plays in theories of consumption and demand. In consumer

theory, that is, orthodox economists concentrate on the results

of free consumer choice among a given bundle of commodities

with different prices, rather than focusing on the ways in which

institutions tend to define or limit the bundles available for

choice."

David Gordon, Economic Theories of Poverty and Underemployment (National

Bureau of Economic Research, 1971), p. 55.

I 83

LIIIIINNOMICINIMIMINIIIIIIIIIIIIEMIIiii.

66

stratification model of earnings, individual choice or "free will"

is consigned to the error term, as it were, along with other

stochastic variables.

The distribution of earnings is a function of successive

stratification in three markets. In the human capital market, race,

sex, and social cla-is, or what may be called the "social stratifi

cation" factors, play a predominant role in the distribution of

personal investment opportunities. In the "external" labor market,

race and sex play an essential role in the distribution of individuals

across occupations and industries, human capital held constant.

Finally, in the "internal" labor market, these same social stratifi

cation factors are responsible for part of the variance in earnings

within specific occupations and industries, again assuming human

capital ceteris paribus.

The General Social Stratification Model

A simplified version of the stratification theory can be

described in a recursive system of functional equations. Race, sex,

and social class are the ultimate exogenous variables which determine

the earnings distribution through a series of primary and supplementary

transformations on human capital, occupation strata, and industry.

(I) HCi+ F

HC(R

i'Si'

Ci.,

Ai) + eHC

(II) OSik = FOS(HCi' Ai) + fos(Ri, Si, Ci) + cos

(III) INij

= FIN

(Ri

, Si) + EIN

i 84

67

(IV) Wi = FW (OSik'

INij

) + fW(HC

i'Ai) + fiW (R

i'Si) + E

w

where: HC = human capitalR = race

:; = sex

C = social classA = "innate ability"

OS = occupation stratumIN = industryW = wage ratei relates to the i

thindividual

.thij relates to the Jo, industry

k relates to the k occupation

F is a primary functionf is a se,_ondary function

f' is a tiertiary functionE is a residucal term

Equation (I) indicates the expected maximum quantity of human

capital which individual i will be able to acquire.2

Following

Becker, the quantity of human capital demanded is positively corre-

lated with ability.3 Ceteris paribus, those with greater native

ability have higher marginal rates of return (mrr) on any given level

of investment and thus have an incentive to invest more than others.

27f the present model is used to evaluate the present value

of discounted lifetime earnings, HC would refer to the expectedlifetime acquisition of human capital. If the model is used to

evaluate wage rates at a given point in time, HC ls a measure ofthe individual's human capital at the time when W is measured. To

the extent that individuals halo. different time paths of capitalacquisition, the distribution of earnings at a point in time will

diverge from the distribution of lifetime earnings.

3See Gary Becker, "Human Capital and the Personal Distributionof Income: An Analytical. Approach," W.S. Woytinsky Lecture No. 1,

Institute of Public Administration and Department of Economics,University of Michigan, 1967, p. 5.

85

68

The available supply of human capital, however, is restricted

by racial, sexual, and class discrimination in the capiLal funds

market while the demand for funds is limited by discrimination in the

labor market. In the capital market, minorities find the marginal

cost of investment funds (mcf) higher than for others. This affects

the quantity of personal human capital accumulated as well as its

structure. Entry barriers to apprenticeship programs, for example,

are equivalent to prohibitively high mcf rates for specific types of

investment.

On the demand side, external and internal labor market

discrimiacion diminish the equilibrium marginal rate of return on

investment for minorities by lowering future expected earnings.

Together, the lower mrr on investment and the higher marginal cost

of funds constrain the amount of human capital acquired by minority

members of the labor force. Human capital acquisition, according to

the stratification theory, is thus a function of innate ability

tightly constrained by the onus of race, sex, and social class origin.

As suggested earlier, the error term (eHC

) includes the effec. of

personal preference and the rational response to wage differentials

insofar as the individual is net completely constrained by other

variables in the function. This equation is less mechanistic than

may at first appear. The effect of rlce, sex, and social class

operates through cultural transference as well as through institu-

tional discrimination. Social class, for instance, obviously plays

a significant role in determining human capital investment decisions

f

VW

69

by structuring "personal preference." The same can certainly be said

for sex and probably for race. Given this perspective, "individual

decision - making" is for the rust part socially determined leaving

only a small residual which can be thought of as pure individual

personal preference. The actual "size" of this residual, of course,

is open to considerable debate.4

The second equation maintains that the probability of individual

i entering occupation stratum k is determined primarily by the

individual's stock and structure of human capital and native ability.

For present purposes, the concept of "occupation stratum" need not

be rigorously defined.5

Equation (II) can be thought of as the human capital equation

in the overall theory. At this stage, the social stratification

factors enter the equation indepeneAtly, but are of secondary

importance. Their primary role is played in the first and third

equations; that is, minority members of the labor force are assumed

to be E:reened out of certain occupations not so much because of direct

occupational entry barriers, but because of the dynamics represented

4 For an excellent discussion of how social class, family, and

school interact to determine :::e level of an individual's human capital

stock, see Samuel Bowles, "Unequal Education and the Reproduction of

the Social Division of Labor," Review of Radical Political Economics,

Fall-Winter 1971.

5For empirical purposes, an occupation stratum will lath be

defined as a set of specific census occupations which share similar

specific vocational preparation (SVP) and general educational develop-

ment (GED) requirements as reported by the U.S. Departm2nt of Labor,

The Dictionary of Occupational Titles (Washington: U.S. Government

erinting Office, 1966).

1 87

70

in Equation (I). The stochastic term (ens) accounts in part for

Personal, occupation preference after controlling for human capital

and the influence of race, sex, and class.

Unlike occupation strata, industry attachment is based on

race and sex alone (plus a stochastic factor). This formulation

follows from the fact that most industries require a broad range of

skills and combine a large number of occupations. For the purpose of

the model, the whole spectrum of occupation strata in the economy can

be thought of as being replicated in each industry, although the

relative number in each occupation stratum varies considerably. The

theory maintains that there are racial and sexual barriers which

prevent large numbers of minority members from entering certain

industries even in occupations which require relatively little human

capital or innate ability.

The error term in Equation (III) contains a n"mber of factors

beside personal preference. Limits to geographical mobility between

labor markets has some effect on constraining "industry choice,"

given regional differences in industrial structure. Cyclical factors

in the aggregate economy also affect the relative availability of

positions in different industries. In addition, pure "luck" plays

a role in industry attachment; being in she "right" personnel office

at the "right" moment may be an important factor in determining an

individual's attachment to an industry sector.

Finally in Equation (IV), the personal distribution of earnings

is described by the distribution of the labor force into occupation

88

71

and industry "slots." Knowledge of an individual's occupation stratum

and industry attachment is sufficient to define the individual's

wage within rather narrow limits. At this point, differences in

human capital and ability as well as differences in race and sex may

still have an independent effect in terms of further defining

individual earnings.

To summarize, stratification plays its primary role in determining

tne distribution of human capital. (See Figure 3.1) But it

continues to play an independent and supplementary role at every stage

Figure 3.1 A social stratification model of the

personal earnings eistributicm.

in the wage model. Race, sex, and class affect occupational

attachment independent of their effect on capital while race

and sex are also key determinants of industry attachment. Finally

89

72

these same factors, according to the theory, affect the final

distribution of wages through pure wage discriminaticn within

specific occupations and specific industries even when human capital

and ability are equal among workers.

Like all general theories, the stratification theory cannot

explain all of the variance in the earnings distribution. The error

term in Equation (IV) must account for a large number of influences

which may have only the most tenuous connection to race, sex, and

social class. To be a complete theory of wage determination, this

framework would have to be expanded in Lwo directions. First, some

attempt would be necessary to explain why stratification and discrimi-

nation play such a crucial role in the earnings distribution, if

empirically they do.6 And second, some hypothesis would be required

about the demand side of the labor market in order co explain what

appears to be a continuing'disequilibrium in terms of industry "ability

6Why labor market discrimination persists in light of itssupposed negative effect on efficiency and profits continues to be

one of the critical unanswered questIzns in modern economics. Whether

discrimination oczurs because of employer and employ ! "tastes" as

in Becker's early analysis, or discrimination is a rational statistical

response to labor market information costs as in Arrow's treatment,or whether it occ.irs because of "capitalist attempts to divide andconquer the labor force" as in some of tne radical literature cannotbe directly tested here. What can be tested is how powerful strati-fication is it terms of the earnings distribution. For backgroundmaterial on the debate, see Gary Becker, The Economics of Discrimination(Chicago: University of Chicago Press, 1957); Kenneth Arrow, "SomeModels of Racial Discrimination in the Labor Market," RAND PublicationAM-6253-RC, February 1971; and David M. Gordon, Richard C. Edwards,and Michael Reich, "Labor Market Segmentation in American Capitalism,"

mimeo, March 1973.

90

73

to pay."7 Neither of these massive efforts is undertaken here.

Rather a more specific earnings generating function is derived which

can test for the size effect of stratification on personal earnings.

The Specific Model

The general model provides a L'anework for analyzing the total

effect of capital, labor, and product market imperfections on the

distribution of wage income. However, the scope of the present study

is limited to an investigation of only one kind of imperfection.

Here we are concern2d with the extent to which barriers to occupational

and industrial access distort the wage distribution among individuals

of equal human capital endowments. For the sake of the present inquiry,

human capital acquisition is considered exogenously determined. Thus

empirical tests will be primarily restricted to Equation (IV). The

specific model is derived from the "crowding" hypothesis first

explicitly formulated by Edgeworth in 1922 and since rejuvenated by

Bergmann.8

To begin, assume a world in which there are two industrial (or

occupational) sectors and labor is homogeneous in endogenous

7One tack taken to understand the differential "ability to pay"

begins with a theory of uneven development within a dual economy.

For more on this subject, see Robert T. Averitt, The Du.l Economy (New

York: W.W. Norton, 1968) and Barry Bluestone, "Economic Crises and

the Law of Uneven Development," Politics and Society, Fall 1977.

8See F.Y. Edgeworth, "Equal Pay to Men an' Women," Economic

Journal, December 1922 and Barbara Bergmann, "The Effect on White

Incomes of Discriminat-on in Employment," Journal of Political Economy,

March/April 1971.

91

74

productivity and inelastic in supply. Furthermore assume that in

the simplest case demand conditions are identical in both sectors

so that the marginal revenue product curves are the same in Sector A

and Sector B. (See Figure 3.2) If there are no barriers to

$wage

Sector B Sector A

MRPB

MRPA

Employment in B

Figure 3.2 No "Crowding"

AEmployment in A

intersector mobility, in equilibrium an equal number of workers will

be found in both sectors (E*A= E*) and the universal market wage will

be w* = MRP*. Each worker is paid his marginal product which reflects

his endogenous productivity. If we relax the assumption of identical

MRP curves, wages will still be equal assuming a perfectly competitive

labor market.

We can now posit that for some reason firms in Sector A refuse

to employ minority workers, restricting their workforces exclusively

to white men. All other workers are forced to find employment in

Sector B. Assuming that labor force participation does not change

92

75

after segregation is imposed, the resulting wage and employment

relati4 's will be as described in Figure 3.3.

MRPB

Swage

Sector B Sector A

MRPA

Employment InE*A

Employment in A

Figure 3.3 Simple "Crowding"

The total labor force E'A ElBequals the old level EAEB, but the

imposed segregateddistribution of workers creates a wage differ-

ential of (wA-w

B). Each worker continues to be paid the marginal

product in his sector (wA = MRPA; wB = MRPB), but now there is no

correlation between endogenous productivity and relative earnings.

In Sector A, white males are paid a wage greater than their endogenous

productivity would warrant (WA > MRP* = ERP*) while in Sector B, all

minority members are paid a wage below the level that would exist in

a non-segregated economy.In this case we can say That minority

workers are "crowded" into Sector B, resulting in lower earnings.

Imperfect mobility between sectors results in a quasi-equilibrium

where wage differentials can persist and where total output is below

93

76

its full equilibrium level.9

Once crowding exists, differences in the labor demand schedules

of the two sectors can affect the earnings differential. In

Figure 3.4, Sector A is drawn so that the marginal revenue product

MRPB r" i

1e

s

s

Employment in B

$wage

wA

t::a

wB

MRP'A

MRPA

1i

EA EA-'.Employment in A

Figure 3.4 Complex "Crowding"

of labor is higher than in sector B for every equal level of employ

ment. EAEB

continues to represent the total supply of labor

(=EIEt), while minorities are limited to employment in Sector B.

Under these conditions, the wage differential will be larger. Either

because of higher marginal physical product (MPP) or higher marginal

revenue (MR) or both, workers who have access to Sector A will benefit.

9Obviously this result requires imperfection in the product market

as well. If all product markets were perfectly coapetitive, any

employer who paid a wage higher than MRP* to attract a full complement

of white male labor would shortly be forced out of the market. At a

minimur this model requires some imperfection between economic sectors.

1 94

77

If for some reason minority workers were restricted to the sector

with a higher marginal product, it is possible that demand conditions

could offset the observed effect of simple crowding. We expect that

in most cases, however, minorities will be crowded into those sectors

where demand conditions are relatively weak, thus adversely affecting

their relative wage positior. Over time there is a tendency for

simple crowding to become "complex." Industries in sector A will tend

to substitute capital for labor because wages are high, while

industries in B will substitute labor for capital. The present situ-

ation in U.S. labor markets may reflect this long-run effect.

There is no problem in generalizing this model to n sectors.

Assuming homogeneous human capital, imperfections in the product

market, and the existence of crowding, the complete earnings distri-

bution would be described by the set of quasi-equilibrium wage rates

established in each sector. Nor is there a need to speci7y perfect

segregation by race, sex, or some other non-endogenous productivity

factor for the crowding model to be perfectly serviceable. One of

the key hypotheses to be tested, in fact, is that the distribution of

earnings is a function of the degree of crowding in each occupation

and industry. Ceteris paribus, the smaller the proportion of

minorities employed in an occupation or industry. the higher the wage.

A Mathematical Treatment

The crowding hypothesis can alEo be described mathematically.

In doing so the parameters that determine wages in the presence of

market segmentation can be derived. Assume once again that human

i 95

78

capital Is homogeneous in a two sector labor market. Furthermore

assume that the marginal productivity of labor i3 a linear function

of the number of workers in each sector and is independent of the

number of workers employed in the other sector. Finally, assume that

employers are unwilling to pay a wage to each worker that exceeds

marginal productivity and that workers in a given sector refuse

employment which fails to pay them their marginal product. This

assures that sectoral wage rates will never be above nor below the

marginal revenue product in each respective sector. These last

assumptions are only included to simplify the exposition.

The model can be expressed in two linear labor demand equations:

(1) wA = aA - bAEA

(2) wB = aB - bBEB

where b = dw/dE

and one employment constraint:

(3) ET = EA + EB.

By mlking alternative assumptions about the intercept term in

sector, a., the relative slopes of the MRP curves, bi, and the

number of workers employed in each sector, Ei (determined exogenously),

measures of the wage differential between the two sectors (6 = wA - wB)

can be derived.

Four different cases of crowding can be isolated.

96

79

(I) No Crowding (with identical demand curves)

Assume aA = aB

bA = bB

ET

is mobile between sectors

In this case there are no barriers to intersectoral mobility. Any

wage differential between the two sectors will induce some workers

to move from the lower wage sector to the higher wage sector until

wage rates are equalized throughout the whole economy. In this

instance, because the demand curves are identical, employment will

be divided equally between the two sectors in equilibrium (EA = EB)

and

(4) 6 = wA - wB = (aA

- bAEA) - (aB- bBEB) = 0

(II) Simple Crowding

Assume aA = aB

bA = bB

EB > EA

Here the MRP curves are identical, but minority workers are excluded

from Sector A. Therefore,

(5) 6 (aA bAEA) (aB bBEB)

= b(EB- E

A)

_ dw(E - EA)

dE B A

97

80

In this case the total wage differential is a function of the number

of workers confined to each sector. The greater the slope of the

identical demand curves, the larger the wage differential given

EA and EB.

(III) "Complex Crowding" - Type 1

Assume aA > aB

bA = bB

EB > EA

In this case minorities are crowded into Sector B and the demand

curve in Sector A is above the schedule in Sector B. At every level

of equal employment in both sectors, the marginal revenue product in

A is greater than in B. The wage differential, 6, will then reflect

both the "supply" effect of segregation and the "demand" effect of

the vertically shifted MRP curve.

(6) 6 = (aA - aB) + b(EB - EA)

dw= (a -a ) + [--

dE(E

B- E )i

A B

In the linear model, the two effects are simply additive although

the existence of segregation is a necessary condition for the

existence of any "demand" effect.

, 98

81

(IV) "Complex Crowding" - Type 2

Assume aB > aB

bB > bA

EB > EA

In this more general case, the demand curves have different slopes

as well as different intercepts. Here MRPB has a lower intercept

and is more inelastic than MRPAat every level of equal employment

in the two sectors.

(7) 6 = (aA - aB) + (bBEB bAEA)

dw dw(aA aB) I[dEl EB (dEl

A"A]

B

This last equation can be expressed in terms of demand elasticities

by substituting (wi/Ei)/fli for the bi in each equation. Therefore,

WA148 (11 E -[(-21 n(-±-) EA(8) 6 = (a a ) EB n B EA AA B

Rearranging the terms in equation (8) yields:

(8') wA

(1 + )- w (1 + --1--)= (aA

- aB).:I

AB riB

1 99

82

(8")

wA

(aA

- aB) (1 + 1/n

B)

wB wB (1 + 1/nA

) (1 + 1/n A)

Finally, letting wB be the numeraire (wB = 1), we can derive an

expression for the relative wage, w;i.

1 + a.A

- aB+ 1/n

B(9) w'A 1 + iin

A

Thus with employment levels set exogenously, relative wages

will be a function of the loci of the respective sectoral demand

curves. More specifically, if the intercepts are equal (aA = ad,

expression (8") reduces to:

wA

(1 + 1/n )B

(10)wB

(1 + 1/nA)

and it is clear that, given intersectoral immobility, relative wages

are a function of relative employment levels and the labor demand

elasticities in each sector.

One interesting implication of the "complex crowding" model is

that in the face of intersectoral immobility, the earnings of

minorities may still be equal to or even exceed those of the dominant

employment group if the labor demand schedule in the crowded sector

is sufficiently above that in the discriminating sector. From

equation (7) it is clear that: given equal intercepts, the wage

100

83

differential is reduced to zero when the ratio of the demand slopes

is inversely proportional to the ratio of employment in the two

sectors. That is, 6=0 when be /bA = EA/EB and alt. aB. More

generally, when the intercepts are not equal, 6=0 when:

1(11) n -

B 1/nA+ aB - aA

This follows from equation (9).

While this may be only of theoretical interest, it implies that

if for some reason crowding could not be overcome, wage equalization

could still be brought about by manipulation of the derived demand

for labor in each sector of the economy. That is, if somehow the

demand curve in the crowded sector can be raised above the demand

schedule in the discriminating sector, .the wage differential can be

reduced. Increased demand in the crowded sector can thus compensate

for the earnings effect of "oversupply."

The Reduced Form

To measure the composite effect of "crowding" and differentiated

labor demand conditions, it is necessary to hold endogenous productivity

characteristics constant and investigate the remaining variance in the

earnings distribution. This is equivalent to standardizing for human

capital and then carefully measuring the composition of the remaining

wage differential, 6.

1 101

84

Assuming that endogenous productivity is measured perfectly,

a non-zero differential indicates either some degree of crowding or

an inequality between wage rates and marginal revenue products, or

in all probability, some conbination of the two. The portion of the

differential due to the relative positions of the employment suppy

curves, EA and EB, and the MRP schedules can be identified as resulting

from industry or occupational crowding. Any remaining differential

must then be due to either imperfections in labor market information

or to pure wage discrimination within industries and occupations

(assuming no measurement or specification error in the human capital

and "crowding" variables).

Obviously this course of empirical investigation is fraught

with obstacles. Controlling totally for endogenous productivity is

an impossibly difficult task. There are a myriad of individual

characteristics which enter into the composition of an individual's

endogenous productivity. Measuring even a small number of these,

independent of the price they exact in the market, requires careful

specification. Even then it is difficult to know how much of the

remaining variance may be due to unmeasured endogenous productivity

traits.10

10Is physical height, for instance, an important "endogenous

productivity characteristics" for salesmen? If it is and this par-

ticular variable is not included in the earnings generating function,

we will obviously fail to account for all the variance in salesmen's

salaries. Worse yet we may erroneously attribute some of the variance

in earnings to another variable which is covariant with height. In

this case we run the risk of.fostering a mistaken conception about

the arguments in the earnings function.

102

85

A no less difficult problem arises in identifying the labor

supply and demand schedules for each sector of the economy. Measuring

the elasticity of demand for labor in a particular industry, let alone

in all sectors of the economy, poses some severe methodological

problems. The same can be said for measuring the sectoral labor supply

curves, even accepting the simplifying assumption of zero elasticity.

Finding useful proxies for identifying the loci of the individual

supply and demand curves consequently requires some ingenuity.

Still another problem arises in specifying the functional form

of the final model. In anything so complex as wage determination,

many factors will enter interactively rather than independently.

However the relatively simple substitution of log space for even

simpler linear space may add very little power to an earnings

generating function; the actual interactions between variables may be

much more complex than log linear.11

11The size distribution of personal income in most Western

capitalist nations appears to be lognormal leptokurtic with a Pareto

upper tail. Consequently, in order to explain how this distribution

occurred, many investigators have attempted to replicate this form

through variations in a lognormal function of human capital factors.

This research has had mixed results. See, for instance, Lester C.

Thurow, Poverty and Discrimination, op. cit. In this work, Thurow

fits an equation of the following form:

cIik

= AEd.bExk

where Iik

= income for an individual with i years of education and k

years of experience; A = shift coefficient; EDi = i years of

education; Exk = k years of experience; and b and c are income

elasticities. He concludes that education interacting with years of

work experience is an important ingredient in explaining the

103

86

For all practical purposes, it is impossible to deal

definitively with these problems in an empirical analysis. It is

possible, however, to specify some of the most obviously important

human capital variables and then, using a number of carefully con-

structed industry and occupation indices, investigate the extent to

which the remaining variance in earnings can be explained by market

imperfections. Two stage equations might be used for such an analysis,

assuming that an individual's endowment of native ability and his

acquisition of human capital temporally precede his entry into a

specific occupation and industry. In this case the first equation

would specify individual earnings as some function of endogenous

productivity characteristics plus a residual term, el. The second

equation would attempt to explain El in terms of industry and occupation

variables acting as proxies for measures of labor supply and demand.

Following this procedure and assuming careful measurement of

all variables, there would be strong evidence in support of the "pure"

human capital theory if the first equation accounted for a large part

of the variance in earnings while the second equation failed to explain

much of the variation in El. Conversely, if a large portion of the

variance in earnings was explained by equation two, this would consti-

tute evidence of significant labor market imperfections. The substance

distribution of earnings. For a more general theory of comple-

mentarities among independent variables in income generating functions,

see Martin Bronfenbrenner, Income Distribution Theory (Chicago:

Aldine-Atherton, 1971), pp. 50-54.

104

87

of these imperfections could only be known if one were to have some

confidence in the proxy variables for segregation. If these variables

truly measure "oversupply" or differential demand, then significant

coefficients in the second equation are a strong indication of

"crowding" and the residual in this equation, E2, measures, at least

in part, the earnings effect of imperfections in labor market informa-

tifm and/or "pure" wage discrimination. Thus in a well-specified

system of equations it would be possible to measure (1) the effect 'If

differential endogenous productivity on the distribution of earnings;

(2) the effect of industry and occupation crowding on wage differ-

entials between individuals; and (3) the residual effect of information

imperfections and pure wage discrimination.

No matter the propriety of a two stage analysis for testing the

stratification theory, a single equation reduced form has been used

in the present research. The regression equations take the familiar

form:

Wi

= cti. + Er3ixi + E

The use of this ec ition is warranted by the relative intractability

of more complex equation systems and by the prohibitive cost involved

in actually fitting large amounts of micro data in multiple stages.

This procedure Is not unusual in that virtually all previous attempts

at measuring the determinants of earnings through large micro samples

105

88

have also relied on single regression equations.12

'13

For the same

reasons of tractability and cost, the basic equation is fundamentally

additive.14

The right side of the equation is composed of four groups or

"modules"ofX.variables. One controls for human capital; another

controls for non-monetary effects on relative wages due to working

conditions; and the last rwo are proxies for the loci of the labor

supply and demand schedules. The actual regression equations take the

linear form:

12See, for instance, Morgan, et al., Income and Welfare in the

United States, op. cit.; Weiss, "Concentration and Labor Earnings,"op. cit.; Rees and Shultz, Workers and Wages in an Urban Labor Market,

op. cit.; Stafford, "Concentration and Labor Earings: A Comment," oz.

cit.; Johnson and Youmans, "Union Relative Wage Effects by Age andEducation," op. cit.; Bennett Harrison, Education, Training, and theUrban Ghetto (Baltimore: John Hopkins Press, 1972); and Wachtel andBetsey, "Employment at Low Wages," op. cit.

13Actually, as explained later in the text, a decision rule was

followed in fitting the equations such that an approximation to asequential equations model is obtained. In effect, the human capital

variables are held constant (or nearly so) when the industry andoccupation variables are added. This is analogous to allowing theselatter variables to explain only the residual variance in the earnings

function.

141n running the regressions, several log linear forms were

tried on several sets of data. In each case, the log transformequations did not perform appreciably better than the simpler linearequations, and in a few cases they performed a bit worse. For this

reason, and also because the additive model was much easier to inter-pret and evaluate, the final regressions were run in the additiverather than interactive form. In future research I hope to experiment

with several different transformations on the raw data. These may yield

somewhat better results if a transform can be found which more approxi-

mates the actual underlying interactions between independent variables

and the true relation between independent and dependent variables.

106

89

wij.. = a +71bj HC +1:b INDijrskrs ijkrs jmrs mrs

k m

+12:b.3sSTRAT +Ebj WC + e

ijnrs prs ijprs

n p

where wijrs

= wage rate for individual i of race r and sex s

in occupation stratum j

HC = human capital characteristic k for individual iikrs

IND.. = industry characteristic m associated with theijmrs industry within which individual i is employed

STRAT.. = a measure n of industry or occupation "crowding"ijnrs for the industry or occupation within which

individual i is employed

WC.. = a measure p of working conditions in the specificijprs occupation within which individual i is employed.

= an error term

The ability to accurately estimate this set of equations depends

on the existence of a suitably large comprehensive micro data set

and an adequate specification of each module.

107

90

The Data Source and the Set ofRegression Variables15

The basic data for this study is taken from the 1967 Survey of

Economic Opportunity compiled by the Office of Economic Opportunity and

the Bureau of the Census.16

A total of some 61,000 individuals are

found in the SEO file, approximately half of which are contained in a

self-weighting sample of the United States population. The other half

of the sample is drawn from individuals living in predominantly

nonwhite census tracts. This oversample provides much better estimates

of nonwhite population parameters and consequently it is used along

with the blacks in the self-weighting sample to estimate the black male

and black female equations.17,18

15For an extended description of the data base and how it was

compiled see Appendix A.

16The Survey of Economic Opportunity is available from the SEO

Clearinghouse, Data and Compilation Center, Social Science Building,University of Wisconsin, Madison, Wisconsin 53706. More informationabout the SEO can be obtained from the Clearinghouse including codebookand user's guide.

17Comparisons of the means and standard deviations from the

nonwhite segment in the self-weighting sample and the nonwhites in thespecial oversample indicated no significant differences in terms ofall of the variables used in. this study. However, there were signi-ficant differences between the whites in the selfweighting sampleand the whites included in the oversample. For this reason, theoversample population was added only to the black equations. The

N's were already of sufficient size in the white equations and theaddition of this special sample to the black equations allowedextensive stratification of the black population without loss ofstatistical significance. The oversample is not used in the race-sexpooled regressions.

18Unfortunately, it was necessary to delete nonblack nonwhites

from the sample population. The SEO does not contain large enough

108

91

From the SE0 file, all full-time, full-year workers were

selected.19

This subsample was further refined by the elimination of

all those who either did not report a wage rate or reported that their

present job s not their "usual job." In addit corkers below

age 25 were excluded leaving a sample of predominantly prime age

individuals. The total N in the final sample is 13,896.

Data on specific occupation, specific industry, race, sex,

hourly wage rate, years of schooling completed, region at age 16,

migration from place of residence at age 16, and union membership status

were obtained for each individual from the 1967 survey. In addition

data on vocational training were available for nearly three-fourths of

the sample from the 1966 SE0 panel. Where the 1966 and 1967 SE0

individuals matched, their training data was merged onto the 1967 tape.

Industry and occupation characteristics available from a number

of macro data sources were then merged onto each individual record in

the sample. Thus each final record contained not only data on an

samples of other minorities to oermit statistical analysis. At the

same time, other minorities h- ! sufficiently different labor market

experiences that to include them with blacks would bias the empirical

results. For information on different labor market experiences ofminority groups, see Larry Sawyers, "The Labor Force Participation ofthe Urban Poor," Ph.D. dissertation, University of Mtchigan, 1969.

19Full-time, full-year represents all those who (1) reported 30or more hours of work in the week preceding the interview and(2) reported 40 or more weeks of employment in 1966. This definitionis somewhat more lenient than the normal Labor Department definition.It was used in order to take into account those who have normal full-time jobs with some degree of seasonality and those who have full-yearjobs where a full work week is somewhat less than a full forty hours,a situation which is becoming more prevalent.

109

92

individual's schooling, for instance, but also on such factors as

the profit rate and the concentration ratio in the specific three-

digit SIC industry in which the individual worked in 1967.

After merging the macro and micro data, the total sample was

stratified into occupation groups. Each of the 298 census occupations

was matched to the Dictionary of Occupational Titles yielding

unweighted average General Educational Development (GED) and average

Specific Vocational Preparation (SVP) scores for each occupation.20

From these scores, seventeen occupation gtcd.,is were formed which

were ordinally ranked according to GED and SVP. Next, in order to

create strata with sufficient sample size, groups were added together

to form five broad occupational strata. These are the strata used in

the final analysis. Each stratum contains occupations with the same

narrow GED range and (except for stratum 5) a broader range of SIP

scores. The final occupation strata include groups 1-3, 5, 6-9, 12-14,

and 15-17. Occupation group 4 was too small to be included in the

study. Occupation groups 10 and 11 include "clerical and kindred

workers, nec" and "salesmen and sales clerks, nec." Because of the

heterogeneous nature of these categories it was necessary to eliminate

them from the final analysis.21

20For detail on the Dictionary of Occupation Titles and theconstruction of the GED and SVP scores, see Appendix A.

21 Some regressions were estimated for occupation groups 10 and

11. Except for a very weak coefficient on years of schooling completed,

there were no significant results and the coefficients of determination

were always below .05.

93

In terms of general occupational descriptions, the strata include

the following types of workers.

Occupation Stratum Type of Workers

1-3 Laborers, unskilled workers, menialservice personnel

5 Operatives, semi-skilled workers, semi-skilled clerical workers, semi-skilledservice personnel

6-9 Skilled operatives, semi-skilledcraftsmen

12-14 Mechanics and technicians, skilledcraftsmen, skilled service personnel,foremen

15-17 Professionals, high-skilledtechnicians, managers, officials

This technique of occupational stratification offers a distinct

advantage over other methods of categorizing the labor force.

Ordinarily, workers are classified into one or two-digit census

occupation categories which are differentiated according to job

title rather than the presumed requirements of the job. Following

this procedure, an operative, for example, is never compared with a

given subset of clerical workers or service personnel. Yet for many

operatives, the human capital requirements assumed necessary to

perform a given job with average proficiency are similar to the require-

ments established for workers in some clerical or service positions.

By dividing the sample on the basis of GED and SVP scores rather than

job title, we are able to compare individuals who fill positions having

similar educational and vocational requirements but who are employed

94

in different cenfius-defined occupations. This allows the analysis

of earnings to be carried out for well-defined segments of the

workforce.

The final variable set was chosen from over 180 variables on

human capital, industry, and occupation and represent the closest

proxies which could he found for each of the modules.22

In many cases

specific variables were chosen in order to make the final results

comparable with previous research.

The dependent variable used throughout the analysis is hourly

earnings which is computed in the SEO from weekly earnings and weekly

hours worked. This variable may be biased by differential overtime

rates, but it is still superior to the usual measures of annual

earnings. In most cases, the hourly wage should refer to the

individual's normal wage because of the "usual" job restriction placed

on the sample. Only in the case of abnormal overtime would a problem.

arise.

The independent variables are divided into four modules.

While the modules clearly overlap in some cases, each is an attempt

to measure an identifiable force in the earnings generating function.

HUMAN CAPITAL MODULE The seven factors in the human capital module

are used to measure the effect of acquired endogenous productivity on

individual earnings. In addition, by including these variables in the

22See Appendix A for a discussion of how the data were developed

and a detailed description of each variable.

112

95

regression equation, we can hold them constant and investigate the

effect of other variables consistent with the stratification

hypothesis.

The seven human capital variables include the following:

Schooling - Formal education is measured by the commonly used

variable, years of school completed. This is a continuous variable

with the normal expectation of a positive correlation between it and

the dependent variable. In the linear additive mode, (3 can be

interpreted as the mean marginal hourly earnings expected from an

additional year of schooling.

School-South - To control for the effect of school quality on

earnings, an interaction term is used. The school-south variable

equals the years of schooling completed multiplied by a dummy variable

(=1) if living in the south at age 16. It is expected to be negative.

A

The sum of the as for the schooling and school-south variables yields

the additional earnings from a year of schooling controlled for

region. Clearly this is not an optimal quality control measure for

a number of reasons, but better measures were not available.23

23The inherent problem with the school-south variable is that

it may measure the effect of "region" per se rather than the effect

of school quality. This is particularly true if there is little

interregional migration after age 16. In this case, if the effect of

"region" operates through factors unrelated to human capital, thefinal equation will overestimate the impact of endogenous productivity

on earnings. This will, of course, favor the human capital explanation

of earnings rather than the stratification hypothesis. Ceteris paribus,

the bias in this variable is in favor of the null hypothesis that the

industry and stratification variables have no effect on earnings.

i 113

96

Geographical information on state or SMSA, which would have been

useful for merging educational resource data onto individual records,

was deleted by the Census Bureau from the SEC user tape for reputed

reasons of confidentiality.

Training - Current or previous enrollment in an institutional

manpower training program is measured by a dummy variable.24

This

variable is intended to measure specific vocational training beyond

regular schooling. According to human capital theory, its coefficient

should be posilave representing a financial return to general training.

The coefficient would be zero only if the training was financially

provided by the current employer at no expense to the worker.25

Migration - Geographical mobility is also measured by a dummy

variable which takes on the value of 1 if an individual has not changed

residence by more than 50 miles since age 16. Migration is considered

an investment in human capital insofar as it raises the marginal

product of the migrant.26 Migration, in this sense, is analogous to

24Vocational training covered by this variable includes:(1) business college or technical training (2) apprenticeship training(3) full-time company training (4) vocational training in the armedforces (5) other formal vocational training and (6) non-regular general

schooling.

25For a discussion of the theory behind specific and generaltraining, see Jacob Mincer, "On the Job Training: Costs, Returns, andSome Implications," Journal of Political Economy, Part 2, Supplement:October 1962; and Gary Becker, Human Capital, op. cit. esp. Chapter 2.

26For an excellent discussion of the human capital theory ofmigration, see Samuel Bowles, "Migration as Investment: Empirical Tests

of the Human Investment Approach to Geographical Mobility," Review of

Economics and Statistics, November 1970.

114

97

investment in schooling or training. Given rational mobility, we

expect a negative coefficient on this variable. To the extent that

migration is undertaken for non-monetary reasons or is involuntary,

it is possible that the coefficient will not be significantly

different from zero for some groups.27

Experience - No direct measure of labor force experience is

available in the SEO. As a substitute, a variable which measures

experience as a simple function of age and formal education was

created.28

This assumes that once individuals leave school, they

immediately join the workforce and work continuously thereafter. Because

of the large variance in the pattern of female labor force participation,

this is not a particularly good measure of work experience for women,

yet it may be an adequate proxy for men.

Experience, according to human capital theory, is an important

factor for it is a form of directly usable specific on-the-job training.

The experienced salesman, for example, is more productive because he

not only knows his product, but learns through experience the personal

quirks of his customers. To account for this effect, a number of

27IfIt working married women move in response to the employment

opportunities of their husbands, migration may not have a salutaryeffect on their earnings. Thus the coefficient may very well be zero

for women. This may be complicated by a racial effect for historicallynorthern migration by blacks has been beneficial, no matter the reasonsfor mobility. Thus while white women may not benefit from migration,black women (and all men) might.

28The variable was created by making "experience" = age - years

of schooling completed - 5. This is similar to the construction followedby Thurow and others in creating an "experience" variable for theanalysis of earnings functions.

115

98

previous wage studies have used proxies for experience. Most have

found a strong positive correlation between the proxy and earnings.29

However, the use of "experience" as a human capital variable

is questionable. Because of the ambiguous nature of the "experience"

variable, little can be said about the meaning of a significant

positive coefficient on this factor.30

Nevertheless we have included

the variable in the final regressions and we will normally interpret

it as though its main effect is to augment endogenous productivity.

This, of course, biases upward the total explanatory power of "human

capital" in the earnings generating function. If we were to take the

alternative interpretation of the experience variable--that experience

or seniority reflects nothing more than institutionalized pay incre-

ments based on length of service and set out in collective bargaining

agreements or offered by employers to maintain morale--it would rightly

be considered as one of the industry factors.

29Thurow has used years in the labor force as a proxy for

"experience" or on-the-job training and concludes that a large portionof the difference between white and Negro incomes can be explained bydifferences in the returns to experience. See, Lester Thurow, "TheOccupational Distribution and Returns to Education and Experience forWhites and Negroes," Federal Programs for the Development of HumanResources, Joint Economic Committee (Washington, D.C. 1968), pp. 267-84.Rees and Shultz use "seniority" as a measure of work experience andfind it to be the most significant variable in explaining the wagesof workers in their Chicago labor market study. See Rees and Shultz,

op. cit. In other studies, the variable "age" is often used as anotherproxy for work experience.

30The ambiguous nature of this variable was discussed in the

section on "Empirical Studies of the Human Capital Earnings Function"in Chapter 2.

116

99

Specific Vocational Preparation (SVP) - The final human capital

variable measures the amount of on-the-job training time required to

gain average proficiency in an individual's census occupation. The

actual variable is continuous taking the values of 1.0 through 9.0

reflecting actual training time in months and years. (See Appendix A)

It is used directly as a measure of investment in on-the-job training

and supplements the mcasure of institutional training.

Like the "experience" variable, SVP also has an ambiguous

meaning. It differs from such variables as schooling, migration, and

institutional training in that it does not occur temporally previous

to employment in an industry or an occupation. An individual must

gain access to a specific job before SVP is acquired. Thus if

individuals are barred from entering occupations which require long

training periods, the training may in fact contribute to their marginal

product, but it should not be considered an unambiguous "human capital"

factor. If stratification exists, SVP can be considered an occupation

trait such as union affiliation in a union or closed shop.

Unfortunately there is no independent measure of "native ability"

in the SE0 and consequently the final equations are less than

completely specified according to theory. To the extent that native

ability is positively correlated with acquired human capital, at least

within race and sex groups, the absense of this factor has the effect

of biasing upward the coefficients on the specified variables in the

module. The purely independent effect of native ability must then be

assigned to the error term. A critical problem arises, however, if

1 117

100

innate ability is significantly correlated with industry or occu

pational attachment independent of human capital acquisition. In this

case some of the variance assigned to industry and occupational

stratification in the regression may in fact be due to differences in

ability. While there is no concrete evidence on which to decide this

point, it seems reasonable that innate ability probably has some

independent effect on earnings within an industry or occupation, but

little effect on determining initial employment attachment. The

information costs to the employer of acquiring independent measures of

the native ability of prospective employees probably precludes the use

of such a measure in initial hiring decisions. If this is true, the

effect of native ability on earnings will appear in the error term;

it will not significantly bias the coefficients on the industry and

stratification variables.

STRATIFICATION MODULE For measures of "crowding" or segregation, we

rely on factors which affect the relative labor supply locus for each

industry and occupation. In the stratification theory, these variables

are related to race, sex, and social class. In the traditional

institutional theory, relative supply schedules are determined through

trade unionism and sometimes by other means (e.g. civil service channels).

For present purposes, measures of relative crowding by race,

sex, and union membership status are used. Labor market stratification

occurs along other dimensions as well. However measures of social

class stratification are not available and we can only speculate about

118

i101

other non-human capital characteristics used to segment the labor

force.

Although it is tempting to equate stratification with labor

market discrimination, the hypothesis under examination does not rely

on this interpretation. Stratification may occur through a sociali-

zation process and be related, at least in part, to cultural insti-

tutions and tradition. Women, for instance, may tend to stratify

themselves into certain types of "women's work." Whether this form of

stratification is "voluntary" or not depends on a whole set of subtli

psychological and anthropological questions which cannot be easily

answered.

Union Member - Trade union membership is measured as a dummy

variable for each individual in the sample. Union membership can affect

earnings in two ways; in both cases the primary effect is through the

labor supply schedule. Often in the skilled crafts, labor supply is

directly restricted through apprenticeship programs and work rules

which are maintained so as to limit the number of workers in a specific

occupation. This also appears to occur in a number of professions.

Industrial unionism, on the other hand, has the effect of restricting

employment in a given industry through its influence in setting the

quasi-reservation price of labor.31 In either case monopoly rents are

31 In the case of craft unionism, one can think of the union as

affecting the locus of a perfectly inelastic supply curve of labor,moving it leftward on the horizontal employment axis. In the case of

industrial unionism, the union affects the locus of a perfectly

elastic supply curve, moving it upward on the vertical wage axis. In

119

102

created thus making wages positively correlated with union membership,

given industry demand schedules and of course assuming that unions

are effective in limiting industry employment.32

Percent Minority-Industry (% MININD) - Data on minority

employment was merged onto the SEO sample from the 1960 U.S. Population

Census volume on "Industry Cha-racteristics.-"33

The data refer to

105 three-digit industries. The number of white females plus black

males and black females was calculated as a percentage of total employ-

ment in each industry.

This variable is used as a measure of the relative extent of

segregation in each industry. It implicitly assumes that if there

were no "crowding" there would be an equal percentage of minority

employment in each industry. Industries with relatively few minority

employees are considered relatively "uncrowded." The lack of minority

representation is assumed to be due to some form of entry barrier

which restricts labor supply along racial and sexual lines. The

setting a "minimum" wage below which no labor will be supplied, theunion in effect is setting a reservation price.

32In the long run, of course, trade unionism may also affect

earnings through the capital-labor ratio. Higher wages in the shortrun presents an incentive to the employer to increase the capital-intensity of his production process. In doing so, the marginalproduct of labor is raised, and given labor supply restrictions, thisleads to even higher earnings.

33U.S. Department of Commerce, United States Census of

Population, 1960, "Industry Characteristics," Series PC(2) 7(Washington, U.S. Government Printing Office, 1966).

120

103

percentage of minority employment will then be inversely correlated

with earnings, according to theory.

Unfortunately the mere existence of a significant negative

coefficient of %MININD, no matter how large, is insufficient to prove

the existence of a "crowding" effect. In fact there may not be a

definitive proof of crowding at all because of the difficulty in

isolating this phenomenon from pure wage discrimination.

Theoretically we can distinguish three cases. Pure discrimination

would be the rule if minority workers were paid lower wages in all

sectors while percent minority employment ( %MININD) was invariant.

Alternatively "crowding" would be the best explanation of wage

differences if minorities were segregated into some sectors of the

economy but both minorities and the dominant group (white men) were

paid identical wages whenever both worked in the same sector. Each

of these extreme cases is, of course, clear-cut. Unfortunately the

case which is more realistic is highly ambiguous as "crowding" and

pure wage discrimination probably coexist. It is because of this

"colinearity" that the two independent effects cannot be easily

identified. The best we can do is to amass as much evidence as

possible to draw the distinction knowing full-well that it cannot be

proven. The needed evidence can be gathered by carefully specifying

the estimating equations. This matter is left to Chapter V.

Percent Minority-Occupation ( %MINOCC) - Data on minority

occupational representation was merged onto the SEO sample from the

1960 U.S. Population Census volume on "Occupational

121

104

Characteristics."34

The data refer to 298 specific census occupations.

The number of white females plus black males and black females was

calculated as a percentage of total employment in each occupation.

In addition, variables were created for each minority group separately

as well as one for all females.

Analagous to the industry measure, this variable is intended to

gauge the relative intensity of "crowding" in each occupation. Again

it implicitly assumes that if there were no occupational crowding,

each occupation would have an equal percentage of minority workers.

This variable should be inversely related to earnings, but it has the

same problem of interpretation as %MININD.

INDUSTRY MODULE The five variables in the industry module reflect

an industry's "ability to pay" higher wages. "Ability" is related

to the locus of the labor demand curves in each industry and to the

potential size of producer's surplus. In each case, the variables

chosen relate to the traditional factors used in institutional

analyses of wage differentials.

Concentration (Market Power Factor) (MPF) - The measure of con-

centration used in this study is a new one developed specifically

for merging with the SEC data. Similar to the four-firm or eight-firm

concentration ratios normally used as a proxy for measuring oligopoly

34U.S. Department of Commerce, United States Census of

Population, 1960, "Occupational Characteristics," Series PC(2) 7A(Washington, U.S. Government Printing Office, 1966).

122

105

power, the variable used here is a measure of the share of industry

revenues generated by the firms with the largest assets in the

industry.35

The major difference between this "market power factor"

and normal concentration ratios is that the former has a variant

number, of firms in the "largest asset" category. Normally there are

between three and five firms, but the range for the 105 industries

used in the analysis runs from two to eleven. This is necessitated

by the data source used LO compute the variable.

This does not appear, however, to present a critical problem,

particularly since the simple correlation between Weiss's concentra

tion ratios and the MPF's for the manufacturing sector is .89.

Whatever is lost in terms of the specification is more than compensated

by the fact that the new measure can be calculated for the whole

range of industries, not just manufacturing. In this way the full

variance in "concentration" can be taken into account in the empirical

analysis.

As in most previous studies, a positive relationship between

concentration and earnings is expected. This is particularly true

where workers are organized in strong unions. Collective bargaining

power may allow employees to appropriate a share of oligopoly profits

or gain higher wages at the expense of higher consumer prices. Where

unions are weak or nonexistent, ccncentrated industries may pay higher

35See Appendix A for greater detail on the construction of this

variable.

123

106

wagei anyway in order to forestall union organizing drives or for

purposes of employee morale. Firms in highly competitive industries

are constrained in their "ability to pay" by market forces.

Union X Concentration - An interaction term is used to improve

the specification of the relationship between unionization (which

is in the stratification module), concentration, and earnings. A

negative sign is expected on the interaction term. Unionization and

concentration each affect the wage rate positively. But for a given

level of unionization, the higher the concentration ratio, the lower

the wage rate. This follows from a theory of bargaining power and

spatial limitations to firm entry. The greater the economic and

political power of management, the easier it is for management to

withstand union wage demands. Conversely, where labor is unified and

firms are relatively weak, but spatial entry barriers provide an

appropriate "ability to pay," one expects higher wages. Where strong

unions are up against powerful corporations, the ability to extract

wage increases may be diminished. The former case is often found in

construction and trucking, the latter often in durable manufacturing.36

After-Tax Profit Rate - To measure profitability, an historical

after-tax profit rate (on total assets) was computed for each industry.

36Fot more on the theory of cor.Lentration and unionism, see

Harold M. Levinson, "Unionism, Concentration, and Wage Changes: Towarda Unified Theory," Industrial and Labor Relations Review, January 1967.Weiss was the first to use such an interaction term in regressionanalysis and found a significant negative sign in some -' his equations.See Leonard Weiss, "Concentration and Labor Earnings,' .p. cit.

124

The variable is constructed sc that it measures an average profit

rate for the period 1953-1965. An historical measure is required

for while changes in relative wages may be related to current profits,

it is only logical that wage levels are related to long-run, rather

than short-run, net income.37

Capital/Labor Ratio - The capital/labor ratio is measured by

the dollar amount of depreciable assets per production worker in 1965.

It was calculated by carefully merging data from industry tax recordd

and employment and earnings data. Theoretically, the capital/labor

ratio affects the marginal physical productivity schedule in each

industry through the production function. Assuming the existence of

barriers to labor mobility and wages equal to marginal revenue product

in each sector, the higher the capital/labor ratio, ceteris paribus,

the higher the wage.

Government Demand - Public sector influences on product demand

are measured by the percentage of an industry's output purchased

by all federal, state, and local government agencies. It was computed

from the LILInpulmOutput Matrix for 1958.38 Given the size of

goyernment expenditures and its skewed distribution by industry, it is

37There is a potential simultaneity problem raised by this vari-

able for wage costs are one of the determinants of net income (i.e.

= pQ-wL-rK). To the extent that it exists, however, simultaneitybiases the results in the opposite direction from the positive

coefficient we expect.

38Adapted from the United States Input-Output Matrix-1958.

Wassily W. Leontief, "The Structure of the U.S. Economy," Scientific

American, April 1965.

125

108

theoretically possible for government to appreciably affect the

demand schedule for each industry. In effect, the marginal revenue

schedule may be higher in industries affected by government purchases.

Theoretically, a shift in government expenditures from industry A to

industry B will then affect relative earnings if labor is relatively

immobile.

There is, in addition, another explanation for a positive

coefficient on the government demand variable. The Walsh-Healy Act

and other federal and state legislation provide that government

agencies purchase only from firms which pay the "prevailing" wage or

higher. In doing this, however, the government sector may be responsi-

ble for setting higher wages in those industries where it is a major

consumer. This too would explain a positive relationship between

government demand and earnings indicating that, ceteris paribus,

government-induced employment in the private sector offers higher wages.

WORKING CONDITIONS MODULE Two variables which measure occupational

working conditions were added to the final data set in an attempt to

control for non-monetary pffects on the wage rate. Both variables

were calculated from the Dictionary of Occupational Titles. The

working conditions scores which were added to the data set represent

unweighted averages for the census occupations and were compiled from

the specific titles in a manner similar to the calculation of GED and

SVP scores. Neither is a particularly powerful measure of working

conditions, but represent the best data available at the time of the

109

original analysis.

Physical Demands - The physical demand variable measures the

physical strength required to perform a given specific occupation.

The measure categorizes occupations from "Sedentary" (=1) to "Very

Heavy" (=5) and is represented linearly. If there is, on average,

an aversion to jobs which require heavy physical effort, a positive

relationship between this variable and earnings would be expected.

In the absence of labor market stratification, workers would have to

be compensated with higher earnings in order to perform jobs which

require extraordinary physical effort.

Negative Work Traits - The mean number of adverse working con-

ditions in a specific census occupation is the other variable in this

module. Adverse working conditions refer to extremes of heat and cold,

humidity, noise and vibration, and the existence of physical or mental

hazards on the job including fumes, odors, toxic conditions, dust,

or poor ventilation. The more adverse the conditions of work,

ceteris paribus, the higher the wage necessary to induce workers into

the occupation. The specification of this variable, however, may

preclude its usefulness for one extremely adverse working condition

may require more compensation than several minor ones. Again, the

lack of an alternative data source forced reliance on this measure.

OTHER VARIABLES AND DATA The final two variables used in the analysis

are dummy measures for race and sex. As we mentioned previously,

these variables are used in the cross race-sex equations in order to

1,7

110

distinguish between the effects of "crowding" and other forms of

racial and sexual discrimination in human capital and labor markets.

Care must be taken in interpreting these two variables because of the

difference in the underlying earnings generating functions for each

race-sex group.

The original data set compiled for this analysis of the

stratification theory included virtually hundreds of variables, many

of which were slight variations of the factors included in the final

set. The final variables were selected on the basis of their per-

formance in a large number of macro regression equations. Together

with evidence from previous micro and macro studies of wage

determination, it was possible to arrive at a final set of variables

which reflected all the prime ingredients of an earnings generating

function specified in the general stratification theory. These

-variables were then used in the micro regression equations which will

be presented in Chapter V. But first we must deal with the estimation

procedure.

128

CHAPTER IV

THE ESTIMATION PROCEDURE

Translating the general stratification theory into a particular

reduced form is problematic in itself. Moving the one step further

to fitting actual regression equations poses a number of new diffi-

culties. Before dealing with the empirical results, a brief

discussion of methodology is therefore in order. In this chapter

estimation and testing procedures are developed to circumvent possible

econometric obstacles. One of these concerns the existence of potential

multicollinearity in the exogenous variables. Another is the possi-

bility of specification error.

Potential Multicollinearity

A high degree of multicollinearity is always a potentially

serious ailment in econometric analysis.1 In the present context it

1Farrar and Glauber show clearly Why a high degree of multi-

collinearity poses a serious problem in parameter estimation. In

their words:

"The mathematics, in its brute and tactless way, tells us that

explained variance can be allocated completely arbitrarily

between linearly dependent members of a completely singular set

of variables, and almost arbitrarily between members of an

almost singular set. Alternatively, the large variances on

regression coefficients produced by multicollinear independent

variables indicate, quite properly, the low information content

of observed data, and accordingly, the low quality of resulting

111

1291

112

could in fact be fatal. Linear dependence in the set of explanatory

variables would make it impossible to statistically distinguish

between the effect of the human capital variables and the effect of

the industry and stratification factors on the earnings distribution.

In this case we would be reduced to the very unsatisfactory position

of having to resort to pure a priori reasoning in order to distinguish

between the effects of the two kinds of variables. Regressing

earnings on a nonorthogonal set of independent variables would run

the risk of a serious Type I error in which we might reject a true

hypothesis about human capital or at least seriously underestimate

its impact and seriously overestimate the impact of other factors.

For this reason it is incumbent that we test the degree of collinearity

in the data set and use an estimation procedure which minimizes the

possibility of rejecting valid human capital variables which in theory

temporally precede other factors in determining earnings.

Appendix B reproduces the means, standard deviations, and the

zero-order correlation matrices (XtX) for all of the regressions in

the analysis.2

Each matrix has been analyzed for pairwise linear

parameter estimates. It emphasizes one's inability to dis-tinguish the independent contribution to explained variance ofan explanatory variable that exhibits little or no truly

independent variation."

Donald E. Farrar and Robert R. Glauber, "Multicollinearity inRegression Analysis: The Problem Revisited," Review of Economics andStatistics, February 1967, p. 93.

2We shall use (X

tX) to refer to the zero order correlation

matrix following the notation of Farrar and Glauber. (X X) is the

130

113

dependence according to a standard rule of thumb. In addition, a

stricter test for collinearity based on a modification of Fisher's

2-transformation was used to check for significant non-zero

correlation between paired independent variables.

As an example of the test results for multicollinearity, we can

look at a portion of the (XtX) matrix for white males across all

occupation strata. Table 4.1 is representative of virtually all of

the zero-order correlation matrices used in this analysis. It is

clear that this matrix passes the weak collinearity test specified

by Farrar and Glauber.3 The simple correlations between explanatory

variables never exceed an arbitrary r.. = .8 or .9. This is usually

sufficient to rule out singularity which would be manifest in a near-

zero determinant and the consequent explosion of elements of the

inverse matrix (XtX)

-1. But this weak test would certainly not rule

out the possibility of severe arbitrariness in the coefficients of

the explanatory variables or in the size of their standard errors.

The potential impact of multicollinearity on the final

regression results therefore makes an even stronger test desirable.

Modifying Fisher's z-transformation for the confidence interval of an

estimated correlation coefficient fulfills this need. This simple

algorithm tests for substantial non-zero correlation .4 Each pairwise

cross product matrix normalized (by sample size and standard deviation)

to unit length.

3Ibid., p. 98.

4Using Fisher's 2-transformation, the confidence Interval (z)

131

School

School-S

Migration

01.6

CA)

Experience

aV

SVP

MPF

UNxMPF

Profits

Union

%Min-IND

TABLE 4.1

ZERO-ORDER CORRELATION MATRIX - -WHITE

MALES/ALL

OCCUPATION STRATA

School

School-S

Migration

Experience

SVP

MPF

UNxMPF

Profits

Union

.Min -IND

1.000

.069

-.134

-.486

.329

1.000

-.090

-.074

.032

1.000

.063

-.114

1.000

-.121

1.000

rj ... .146

I.128

-.145

.082

-.217

-.007

-.013

-.053

-.061

J-.119

.030

-.047

.041

.011

tt

.105.

-.037

-.044

.062

.051

1.092

.016

.116

-.100

.003

I-.190

-.026

1.000

.547

.307

.178

-.157

1.000

.266

.782

-.162

1.000

l.199

-.133

1.000

-.120

1.000

HUMAN CAPITAL MODULE

INDUSTRY MODULE

STRAT MODULE

115

sample correlation coefficient was tested to see if it was signifi-

cantly larger than an arbitrarily low .100 at the lower bound in the

95 percent confidence interval.5

For the (XtX) matrix presented in

around a sample correlation coefficient can be calculated as:

z zz - r p

zr

where azr

can be approximated by

1

Zr

where Zr is the z-transformation on the sample correlation

coefficient

z is the z-transformation on the population correlation

coefficient

Zr is the standard deviation of the sample distribution

For the 95 percent confidence interval around the sampler1,,

zp

= zr

(1.96) azr

To modify this formula for use as a test of significant non-zero

correlation, the z-transform for an arbitrarily low correlation, z*,

is substituted for z and Fisher's equation is solved for the Iowa

bound.

z* = z* + (1.96) ar

Using Fisher's transformation table and interpolating, the lower

bound rt. can be calculated. For a fuller discussion of Fisher's

test, see Edward J. Kane, Economic Statistics & Econometrics (New York:

Harper & Row, 1968), pp. 246-47.

5For the purpose of the present analysis, a true population

A.. < .100 was considered a strong indication of linear independence in

133

116

Table 4.1, a sample correlation coefficient, according to this

collinearity test, must exceed .146 for the true population coefficient

to exceed .100 at the 95 percent level.

Applying this procedure to the correlation matrix in Table 4.1

indicates that while there are a number of instances where a sample

coefficient exceeds .146, only two of these involve a correlation between

a human capital variable and a variable in another module. In both

of these cases, the relationship is curiously inverse suggesting that

in the determination of earnings, membership in a trade union may be

a substitute for schooling and on-the-job training rather than itself

being a function of human capital.6 Adding the variable for union

membership to an equation which already includes schooling and SVP will

then not bias human capital coefficients downward.

The same test for collinearity shows some degree of linear

dependence among the variables within the stratification and industry

modules. Union membership, concentration, and after-tax profits are

the explanatory variables. While this figure is purely arbitrary, it

was purposefully set at a low level co assure a strong test of

orthogonality. As it turns out, most of the correlation coefficients

in the (XtX) matrices used in this analysis would pass this ortho-

gonality test even if the pt. were set at an even lower level. Beside

being a strong test, the modification of Fisher's z-transformation

allows a consistent test for multicollinearity throughout the whole

analysis. An rt. was calculated for each (XtX) matrix based on a

pl. = .100 and the individual pairwise sample correlation coefficients

wetle compared with these values.

6This inverse relation is fully consistent with the findings of

Johnson and Youmans in their study of the relative effects of unioni-

zation, age, and education on earnings. See Johnson and Youmans, op. cit_.

134

117

positively correlated with each other while percent minority employ

ment is inversely related to all of these. There is also a degree of

linear dependence in the human capital module. The amount of

collinearity within these mcdules indicates that it is necessary in

some cases to choose subsets of the human capital, industry, and

stratification factors to avoid the purely arbitrary assignment of

explained variance within modules. In running the actual regressions

this was often done. This pattern of linear independence between the

human capital variables and the industry and stratification factors

and some linear dependence within each module is for the most part

repeated in all of the (XtX) matrices in the analysis. IC assures a

minimum of bias in our estimates of each module but indicates that

caution must be used in interpreting the coefficients on individual

variables.

The Estimation Procedure Mechanics

To be even more certain, however, that the small amount of

intermodule collinearity does not bias the empirical results, a two

step estimation procedure was followed in calculating the regressions.

This procedure assures the integrity of the human capital variables.

The same procedure was followed in each complete regression.

The first step in the regression analysis involved running

earnings equations which only contain the human capital variables.

In each case an attempt was made to find a human capital module which

1.35

118

maximized the explained variance.7

The second step in the estimation

procedure entailed adding stratification module variables into the

regt.essiun under the strict proviso that the addition of an explanatory

variable must not destroy the "integrity" of the best fit human capital

equation. If the addition of a given stratification variable made a

human capital factor insignificant or statistically reduced its

regression coefficient significantly, the STRAT variable was removed

from the equation to assure the integrity of the HC module. After

the inclusion of any STRAT variables, the industry and working condition

factors were added again under the same human capital provision. In

this way the assumed causal priority of the human capital variables

is not violated by the effect of possible inter-module collinearity.

In every case individual variables enter the model in a causal order

suggested by Lhe general earnings *_henry.

The initial test for module "integrity" stipulated that the

addition of a STRAT or IND variable must not be allowed to reduce a

previously significant HC variable to statistical insignificance at

the .05 level. With a few important exceptions, whenever the addition

of a STRAT or IND factor wiped out the significance of one or more

human capital variables, the newly added factor was eliminated instead.

This process was necessary in instances where there was a significant

degree of collinearity as measured by the z-transformation test.

7In actuality the "best fit" human capital equation was deemed

to be that one which minimized the standard error of the regression

"tilliate(SEEmird'

106

119

The initial t-test for coefficient integrity assures the

tatistical significance of the human capital variables, but it is

incapable of checking for an absolute change in the size of the

coer-icients after STRAT and IND variables are added. Thus a second

more rigorous test was performed on the human ,:apital coefficients

which entailed applying a standard test statistic for the difference

between two means.8

t'

Estimates of t' were computed for each human capital variable when

there was any doubt about: the size of the regression coefficient in

the complete equation. With the exception of three special instances

which will be discussed in the next chapter, t' was found to be always

well below that necessary to substantiate a significant difference in

coefficients at the 95 percent confidence level. In most cases t < 1

and rarely did it exceed 1.25.

By utilizing the collinearity tests and the two step estimation

procedure, the results from the final single regression equations

approach those that would be obtained from the use of a two-stage

technique. The strict integrity of the human capital module estimates

8From William L. Hays, Statistics for Psychologists (New York:

Holt, Rinehart, and Winston, 1963), pp. 314-19.

137

120

maintained in this way allows a robust, if not overly-conservative,

test of the "crowding" hypothesis.

Problems in Parameter Specification

Controlling for collinearity in the multi-module regression

equations assures that the coefficients on the human capital variables

are not biased by the addition of industry and stratification factors.

But errors in the specification o. Al variable and the form of theA

overall equation could result in poG. estimates of $. Of particular

concern is the specification of the dependent variable and the

absence of non-linear terms and complementarities in the exogenous

variables.

The Dependent Variable - Becker and Chiswick, as well as others

who have studied human capital models, use the natural log of earnings

as the dependent variable when investments are measured in time

equivalents (e.g. years of schooling, experience, training) rather

than dollars.9

In some empirical research a lower coefficient of

determination emerges when earnings rather than the log of earnings

is regressed on schooling and experience. Nevertheless, the dependent

variable in the present analysis is the simple linear term, hourly

earnings. This is consistent with the work of Weiss, Morgan, et al.,

Hanoch, Rees and Shultz, and Wachtel and Betsey.10

9G.S. Becker and B.R. Chiswick, "Education and the Distributionof Earnings," American Economic Review, May 1966.

10See Leonard Weiss, "Concentration and Labor Earnings," op. cit.;

118

I121

Before the final analysis was attempted, a number of preliminary

regressions were prepared on individual occupation strata using the

natural log of earnings as the dependent variable. In these experi-

ments the log specification did not perform significantly better than

a linear specification. Both specifications provided similar

coefficients of determination and standard errors of the regression

estimate. In a few cases the log equations performed a bit worse

than others. For this reason, as well as for ease in evaluating the

final results, the non-log specification was retained. While these

experiments were not performed on all occupation strata or the cross

race-sex equations there does not appear to be any evidence that the

dependent variable is less well specified than in comparable studies.

The superiority of the log specification in some research

viz-a-viz the adequacy of the normal specification in the present

study may be explained by the structure of the respective analyses

and the characteristics of the labor force sample in each study.

The present analysis is primarily carried out within individual

occupation strata rather than across the whole spectrum of occupations

in the economy. It is possible therefore that a linear relationship

exists between earnings and human capital variables within a specific

stratum while the relationship is better represented by a log linear

Morgan, et al., Income and Welfare in the United States, op. cit.;

Giora Hanoch, An Economic Analysis of Earnings and Schooling,"

op. cit.; Rees and Shultz, Workers and Wages in an Urban Labor Market,

op. cit.; and Wachtel and Betsey, "Employment at Low Wages," op. r.it.

139

122

form for the economy as a whole. A move from one stratum to the next

in this case would yield a larger than linear increase in earnings

while increased human capital in any given stratum would yield only

a linear increase in wage. The difference in specification efficiency

might also be due to the fact that the present study is restricted

to full-time, full-year prime age workers who are employed at their

"usual" jobs. The relationship between human capital and earnings

may be linear for this group while the log linear relationship found

in some studies may be a function of differential attachment to the

workforce. Differences in education, for instance, mly have a larger

impact on wages between a part-time worker and a full-time worker

than between workers who share a similar attachment to the labor force.

Non-linearities in the Exogenous Variables - A number of authors

have used non-linear human capital variables to account for the

concave earnings profile normally associated with experience, age, or

seniority.11 Normally this is accomplished by running a linear term

and its square additively; evaluation of the first derivative gives

the extreme value of the function while the second derivative assures

that the extreme value is a maximum. Figure 4.1 indicates how such

a function will often appear. If the actual profile looks like AA,

it is obvious that a linear regression estimate can do little better

11For example, Johnson and Youmans use age and age

2in their

analysis of union relative wage effects by age and education. Johnson

and Youmans, op. cit. Rees and Shultz resort to the natural logarithm

of seniority to better fit this factor in an earnings function.

Rees and Shultz, op. cit.

140

$

EARNINGS

123

YEARS OF-EXPER I ENCE

Figure 4.1 The theoretical relationship

between earnings and experience

than BB unless a quadratic form is used. Obviously BB is a very poor

representation of the true relationship between earnings and experience.

Notwithstanding, no quadratic was used in fitting the final

regression equations in Lhe present analysis. Experiments on pre-

liminary equations indicated that the relationship between experience

and earnings was generally linear for the population under study.

This is not inconsistent with the non-linear profiles of previous

research for the present study sample is composed only of those who

are in prime age and working full-time full-year. This excludes those

under age 25 and for all intents and purposes those who are semi-

retired at age 65. Thus we are attempting to fit only the part of

the curve labeled A'A'. The regression line B'B' performs this task

admirably. The addition of a square or logarithmic term would in this

I 141

124

instance fail to explain any more of the variance in earnings.

It is possible that a non-linear term on SVP would have yielded

marginally better results given the scaling of this variable.

However, if there are significant diminishing returns to longer on-

the-job training, as there is for schooling, a non-linear specification

may not be superior to the one used in the analysis.12

In any case,

the amount of additional variance that might be explained by using

non-linear forms in the human capital module probably does not

seriously affect the final results given the sample population. If

anything, non-specified non-linearities in the industry and strati-

fication modules may bias the results in favor of the relative

strength of the human capital variables viz-a-viz the "crowding"

hypothesis.

Complementarities in the Exogenous Variables - A far more serious

specification error is conceivably introduced by the absence of inter-

active relations in the independent variables. The specification

used in this analysis implicitly assumes that the effect of each of

the explanatory factors is independent of all the others and that

their separate effects are strictly additive.13

'or instance, the

...The research of Giora Hanoch is responsible for identifying

tho diminishing returns to schooling for whites and non-whites in the

North and South. See Giora Hanoch, "An Economic Analysis of Earningsand Schooling," op. cit.

13The one exception to this generalization is the use of an

interaction term to specify the relationship between unionization andconcentration.

142

125

amount of experience is assumed to have no influence on the returns

to schooling and the returns to increasing both schooling and

experience are assumed to be equal to the sum of the separate returns

to increasing each variable independently.

Among others, Thurow believes that complementarities are con-

siderably important in earnings functions.14

He has argued that

Returns are not additive but multiplicative. This may beclearly seen in on-the-job experience and education. The

returns from experience depend partially on the trainee'slevel of formal education. Low education levels make sometypes of training impossible and other types expensive, butas the levels rise, training costs fall and the variety oftraining which can be given expands. These complementaritiesalso work in the opposite direction. Most jobs require someknowledge which is peculiar to the job and is not or cannotbe acquired in school. Education and experience combinedyield larger benefits than the sum of the two.

Ignoring complementarities can consequently lead to biased

estimates for factors which enter wage determination in combination

with others. This is particularly true for equations which cover the

whole occupation spectrum. Within a given occupation stratum, however,

we are in effect holding training levels roughly constant while

observing the returns to education. In this case as Thurow has noted,

the regression estimates of returns to schooling and experience are

valid within each training level.15

Insofar as the primary focus in

14Lester Thurow, Poverty and Discrimination, op. cit., p. 71.

Also Lester Thurow, "The Occupational Distribution of the Returns toEducation and Experience for Whites and Negroes," in Federal Programsfor the Development of Human Resources, A Compendium of Paperssubmitted to the Subcommittee on Economic Progress of the U.S. JointEconomic Committee, 90th. Congress, 2nd. Session (1968) Vol. 1,

pp. 267-84.15

Ibid.

143 i

PAGE 126 OF THIS DOCUMENT WAS MISSINGPRIOR TO ITS BEING SUBMITTED TO THE ERICDOCUMENT REPRODUCTION SERVICE BECAUSE ITWAS UNOBTAINABLE FROM THE SOURCE.

144

127

experience and SVP (on-the-job training) cannot be interpreted

unambiguously as human capital variables (See Chapter II). To tie

them more directly to schooling and institutional training would

seriously jeopardize the interpretation of the whole human capital

module. Furthermore, an extensive amount of expensive experimentation

would be necessary before generating the "correct" form of the inter-

action terms, particularly given the large number of exogenous

variables used in this analysis. For these reasons we have relied for

the most part on a simple linear model to test our theory. Further

research may improve our estimates, but the marginal gain does not

seem to warrant the more than marginal cost of obtaining it.

In conclusion we must use a healthy dose of pure common sense

in evaluating the final regressions. Nevertheless we can be confident

that the problems of multicollinearity and specification error do not

seriously impugn the validity of our findings, especially within

specific occupation groups. As it turns out the actual regression

results tend to be eminently reasonable as will be shown in the next

two chapters.

145

CHAPTER V

THE REGRESSION RESULTS

Having outlined a coherent theory and generated an appropriate

reduced form and a suitable estimation procedure, we are finally in

a position to investigate the empirical results. In this chapter

each of the final regression equations will be separately analyzed.

In the following chapter the regressions will be compared and

evaluated so as to identify what portions of existing wage differ

entials are due to differences in human capital versus differences

resulting from occupational and industrial stratification.

Recalling Chapter III, the reduced form to be tested is of the

general form:

w.. = a +1:b. HC +El), IND..lj TS ijjkrs ijkrs jmrs ijmrs

+Eb.STRAT.. +1:b. WC.. + ejnrs ijnrs jprs ijprs

where i, k, m, and n refer to the ith individual with the k

thhuman

capital trait in the mth industry with n

th degree of crowding and j

refers to occupation stratum, r to race, and s to sex. Individual

equations have been run for each race and sex group for each of the

128

1.4 ft

129

five broad occupation strata. In addition pooled regressions have

been estimated for each occupation stratum across race and sex groups

and for each race-sex group across all occupation strata. rinally a

"grand pooled" regression was couputed for the whole workforce

similar to regressions often found in the literature. Altogether

there are sixty final regressions excluding those where the sample

size is too small to permit statistically significant results.

(I) Stratification by occupation group, race, and sex -

Regressions stratified by j, r, and s are used to generate a series

of distinctive earnings functions for each race-sex group. Each

separate regression is generated on individuals whose particular

occupations share similar educational (GED) and vocational preparation

(SVP) requirements. These equations are especially valuable in

exploring the degree to which wage rates vary within jobs which are

narrowly defined by human capital requirements but potentially differ

in terms of industry characteristics. Accordingly the results can be

usol to evaluate the impact on personal earnings of differences in

industrial and occupational attachment within specific labor market

strata. In addition, by stratifying by occupation group it is

possible to ascertain whether specific variables in the model affect

wage rates differentially as one moves up the occupational hierarchy.

More importantly, in running separate equations for each race-

sex group, one can gather some evidence which can be used to isolate

the impact of crowding from the effect of pure wage discrimination.

A significant negative coefficient of %MININD or 7OMINOCC would be

147

130

prima facie evidence of effective "crowding." It would mean, for

instance, that black males who gained access to white male dominated

industries farLd better than their counterparts in crowded sectors.

The same would be true for a significant coefficient in the white

male equations, the interpretation being that white men who have the

misfortune of being "trapped" in minority-impacted industries bear the

onus of crowding as well. The absence of a significant coefficient

on the STRAT variables in the individual race-sex equations would tend

to weigh against the crowding hypothesis. But the case could not be

closed on this account alone. Evidence from pooled race-sex equations

would not necessarily corroborate this negative finding if the original

STRAT variables in the separate race-sex regressions were insignificant

only because of a lack of variance in these measures. This, of course,

occurs whenever there is perfect or near-perfect labor market

segregation (i.e. apartheid).

(II) Stratification by occupation group across race-sex groups -

Stratifying by race and sex therefore leads to downward biased

estimate:: of the effect of crowding as the degree of crowding

increases beyond some point In the extreme case all differences in

earnings would end up in the constant term or in differences in the

coefficients of other regressors and the impact of industry and

occupational crowding could not be directly tested. To remedy this

potential problem pooled race-sex equations are computed for each

occupation group.

148

131

Unfortunately this solution tends to do the job too well. If

the "crowding" variables are colinear with race and sex--as they

obviously are in the case of perfect segregation--then we would now

find a potential upward bias in the new coefficients. It is possible,

for instance, that earnings differences between race-sex groups are

simply the result of "pure" wage discrimination within each industry

and occupation. Crowding may then exist, but even in its absence

members of minority groups would be paid less.

In the case of perfect segregation it is therefore impossible

to determine whether crowding has anything to do with wage determination

at all. But where there is incomplete segregation--which is the more

usual occurrence--the net impact of crowding can be approximated by

running dummy variables for race and sex in the pooled equations.

Because of multicollinearity problems mentioned in the last chapter,

somewhat arbitrary regression coefficients result, but the final

dummied equations at least put a check on the possibility of

overestimating the independen: effect of industrial and occupational

crowding. The true coefficients on the STRAT variables can then be

expected to lie between the values given in the pooled regressions

with and without the race and sex variables.

(III) Stratification by race and sex across occupation groups -

The equations stratified by j are useful for measuring the effect of

industry and occupational attachment on differential earnings within

narrow GED and SVP ranges. But by their nature these equations will

normally underestimate the full impact of human capital on the total

149

132

distribution of wage rates across the whole occupation spectrum.

Increases in schooling, training, and migration are usually undertaken

to move from one occupation group to a "higher" one. To ascertain

the total human capital effect the regressions must be pooled across

the individual occupation strata. This is the third stage in the

analysis. Again wage equations are generated for each race-sex

group independently in order to account for and measure differences

in the structure of wage generating functions.

(IV) The "Grand Pooled" regressions - The final three equations

are pooled across j, r, and s and are constructed so as to yield

estimates of the full impact of both stratification and human capital

throughout the labor force. Race and sex dummies are added in the

last equation in an attempt to generate an estimate of the net relative

impact of crowding on overall earnings. These final equations must

be treated with all due caution because of the estimation procedure

used. The absence of interaction terms in the human capital module,

the linear form of the dependent variable, and the combining of all

race-sex groups in one equation must be taken into account when

evaluating these results. Nonetheless these last regressions are of

interest particularly when evaluated in light of other findings.

The Regression Results

The regressions presented in this section are the "best fit"

equations consistent with the estimation procedure outlined in the

preceding discussion. The R2s for each regression have been adjusted

150

133

for degrees of freedom and the figures in the parentheses are

t-statistics. The 95 percent confidence level has been used throughout

to measure statistical significance1

The descriptions of each

occupation stratum are based on the mean values for the variables in

the cross ra,a- equations. These can be found in Appendix B. We

begin with the lowest skilled stratum an proceed in steps to an

analysis of occupation stratahaving greater GED and SVP requirements.

OCCUPATION SI TUN -3

Jobs in the least skilled occupation stratum require no more

than a short demonstration period for the typical worker to achieve

average proficiency.2 The average worker in this group has less

than nine years of schooling and only 8 percent have any institutional

training. Yet labor force experience averages over thirty years.

In 1967 a disproportionately large percent of this stratum's workforce

was black (27%) while a full Clird (33%) were women.

Half of the workforce are members of trade unions and are

employed in industries which have on average 36 percent minority

employment. Within specific occupations the percentage of minority

employment is even larger--43.9 percent, approximately half of whom are

1Throughout the analysis there are only a few instances where

a coefficient is presented which does not meet or exceed the .05 level

of significance. These are denoted by an asterisk (*). In most of

these ,,e coefficient is significant at better than the .10 level and

the variable reduces the standard error o. the regression estimate.

In the remaining cases the coefficient is reported for comparative

purposes.

2This figure is based on Interpolation of the SVP scale.

151

134

white women. On average each production worker in this sample had

about $20,000 worth of depreciable assets with which to work,

somewhat less than the amount in other strata. Table 5.1 contains

all of the regression results for this group.

White Males - For white men the "best fit" human capital equation

contains only schooling and the interaction term school-south as

significant variables. Together the t.,:o explain 14 percent of the

variance in earnings which average $2.71 an hour. An additional year

of schooling is valued at 7.8 cents per hour if taken in the non-south.

A year of education in the south, however, adds only one cent to the

wage rate.

The addition of the significant non-human capital variables

increases the corrected coefficient of determination (R2) to .315

and reduces the standard error of the estimate (SEE) to less than

$.74 without significantly altering the coefficients in the human

capital module. Trade union membership adds $.32 to the wage rate

which repfesents a differential of approximately 13 percent over the

wage of non-union workers in this stratum. Industry segregation of the

labor force also affects earnings substantially. Ceteris paribus,

those who become "trapped" :In an industry with minority employment 10

y,reater than "average" earn $.68 less (2 X .1633 X -2.0851) than workers

in industries with minority employment one standard deviation below

the average.3 (Standard deviations are reported in Appendix B.) This

3For consistency throughout the analysis the net effect of

152

135

TABLE 5.1

REGRESSION EQUATIONS:OCCUPATION STRATUM 1-3 BY RACE AND SEX

White Male Black Male White Female Black Female Cross aaeo-Sex

Constant

HUMAN CAPITAL MODULE

2.1321 3.3632 2.1695 1.7986 .9708 2.0723 1.0822 1.8098 1.9190 3.2099

Schooling .0779 .0618 .0' .0601 .0853 .0491 .0669 .0469 .0702 .0455

(3.17) (2.74) (1 I. (4.39) (4.01) (2.65) (4.63) (3.47) (4.13) (3.62)

School-South -.0662 -.0559 -.0 c3 -.0562 --- --- -.0314 -.0157A -.0656 -.0453

(3.56) (3.28) (6.27) (5.20) --- (3.26) (1.73) (5.70) (4.98)

Training --- .4266 .3450 - -- --- .3926 .3341

--- (2.54) (2.56) --- - - --- (2.09) (2.28)

Migration - - -.3436 -.2686 --- --- -.2306 -.2734 -.2011 -.2248

(3.54) (3.37) --- --- (2.78) (3.64) (2.03) (2.92)

Experience

Specific Voc. Prep.

STRATIFICATION MODULEUnion Member .3204 --- .4023 --- .2068 --- .2756 --- .3769

--- (2.45) --- (4.56) --- (2.07) --- (3.57) --- (4.62)

Minority--Industry --- -2.0851 --- -.7213 --- -.8593 --- -1.4104 --- -1.1829

--- (4.85) --- (2.86) --- (2.97) --- (4.21) --- (4.03)

Minority--Occupation --- -.4550 --- --- -1.1527

- (2.09) --- (5.28)

% Black Male--Occupation - ----- - _

X White Female-Occupation ---- - ---

Black Female -- Occupation- - -

INDUSTRY MODULEConcentration --- .9213 --- .3992

--- (5.57) .--- (2.36)

Union x Conc. --- --- - --

After -tax Profit--- 5.0373

--- --- (2.36)

Capital/Labor Ratio- -

Government Demand --- 1.1146 - --

--- (2.11) --- - -

WORKINn CONDITIONSMODULFt

Physical Demands - -- --- --- --- --- -.1687

--- (1.96) --- - --_-_ --- (2.4S)

Negative Work Traits --- -.1324A - --

--- (1.90) --- ---

R2 .142 .315 .200 .494 .204 .511 .220 .402 .187 .521

SEE .8174 .7386 .7527 .6034 .4337 .3511 .4475 .3962 .8091 .6269

MEAN $2.71 $2.71 $2.17 $2.17 $1.74 $1.74 $Y 36 $1.36 $2.27 $2.27

N 138 138 253 253 65 65 137 137 277 777

153

136

is the first evidence of industry "crowding" affecting the distri-

bution of earnings; it will appear many times again. We might add

that this result is clearly in opposition to theories of discrimination

which posit that whites must be paid premium T:ages to work in

industries where they are forced to associate with a large number of

minority group members.4

After the stratification variables are included in the regression,

the industry variables fail to explain any additional variance in

earnings. This is consistent with the "simple crowding" hypothesis

where the locus of the demand curve is uniform across industries but

imperfections exist in the labor supply function. Since we are dealing

with white males in this instance, neither race nor sex is directly

continuous stratification and industry variables is measured over the

range ± one standard deviation (±10) about the means. (We will

refer to this measure as a "one sigma" evaluation.) This measure is

used rather than the traditional elasticity concept because it yields

a more intuitive sense of a variable's impact on earnings. The ±10

evaluation indicates the range in hourly wages earned by homogeneousworkers in industries which differ by ±10 standard deviation in con-centration, profitability, "crowding," etc. By using this type of

measure we are neither focusing on the extreme tails of the distri-

bution nor the infinitesimal marginal effect indicated by elasticities.The overall impact of the stratification and industry variables will

often be of larger magnitude than this, but seldom smaller. For a

normal distribution, two-thirds of all observations lie within la of

the mean; for many other distributions including the "pyramidal," the

uniform, and the bi-modal, a larger percentage of observations liebeyond 10 making our measure somewhat conservative. See Daniel Suits,

Statistics: An Introducti "n to Quantitative Economic Research (Chicago:

Ram: McNally, 1963) , pp. 48-51.

4For a statement of this position see Gary Becker, The Economics

of Discrimination (Chicago: University of Chicago Press, 1957) or

Kenneth Arrow, "Some Models of Racial Discrimination in the Labor

Market," RAND Publication RM-6253-RC, February 1971.

154

137

responsible for the barriers to interindustry mobility required for

a significant STRAT factor. Other factors not specified in the

equation, but probably including imperfections in labor market

information, must account for the industry distribution of white men

in this group. It is also conceivable that "lock-in" effects of

seniority and geographical immobility generate part of the wage

differential.

In the complete equation, the physical demands variable is

significant as well. But its coefficient is negative signifying that

the remuneration for heavier work is lower than for jobs requiring less

physical exertion. This inverse relationship is maintained even after

controlling for occupation stratum (GED and SVP), education, union

membership, and the race-sex composition of each industry. If this

relationship is a valid ind:..cation of the true association between

physical demands and earned income, then workers either prefer heavier

work even at the sacrifice of earnings or some workers become "trapped"

in very low wage laboring jobs and cannot easily escape to other

occupations in this or other strata. Unless we accept the implausible

first implication, this result calls into question the validity of

the "compensatory wage" theory--at least for low-skill work groups.

As it happens, the phys'cal demands variable is seldom significant in

the overall analysis and never in the more skilled occupation strata.

The seemingly counterintuitive conclusion implied by this regression

may be due to measurement error, as we noted In a previous chapter.

155

138

Black Males Within occupation stratum 1-3, schooling has an

identical impact on earnings for both black and white men. Although

the former average almost 1.4 years less schooling than their white

counterparts, an additional year of education for either is worth the

same, 7.8 cents per hour. Only in this stratum and in stratum 6-9

is there no significant difference between coefficients on schooling

for these two groups. In all of the other occupation strata the

partial on schooling is statistically greater for whites.

The equality in dollar returns to education within 0CC STRAT

1-3 would indicate a benign condition if it were not for the fact that

internal rate of return calculations show that extra schooling is not

particularly beneficial for either white or black men as long as they

remain in this stratum. For white men the internal rate of return

based on foregone income opportunity is only 1.5 percent while that

for black men is only a little better than 2 percent given lower

opportunity costs.5

Additional schooling is obviously nct the path to

5The internal rate of return calculations in this analysis are

made according to the usual formula:

nEt

C =(1 + r) t

t=0

where C represents the opportunity cost of an added year of schoolingin terms of foregone earnings, E

trepresents the additional earnings

in period t due to the added year of schooling and r equals theinternal rate of return. In these calculations the opportunity cost,C, was set equal to the annual income earned by an average individualin the occupation strata with mean years of education.

156

139

much higher lifetime incomes at least for those "trapped" within

this stratum.6

As is, the difference in schooling completed can

explain only 20 percent of the difference in mean earnings between

the two groups based on the black male equation evaluated at the mean

value of schooling completed by white men.

As in the case of white men, southern schooling adds little (in

this case, nothing) to the wage rate of black men. But unlike the

results for the white group, both institutional training and migration

have a ?urge impact on earnings. Training adds $.43 to hourly earnings

while the failure to emigrate reduces the wage by over $.34 an hour.

Training apparently permits the black worker to move out of the laborer

occupations (laborers, n.e.c. and farm laborers) into higher paid jobs

in this stratum such as ,arehousemen, metal filers, textile knitters

and loopers; and unskilled painters. Migration represents mobility to

the higher wage labor markets of the north.

The addition of the STRAT and IND modules increases the R2

to

almost .50. Union membership adds over $.40 to the wage rate; thus

the average union member in this stratum earns more than a fifth

The annual additional earnings from an added year of schooling is

assumed to be uniform from the time the individual leaves school until

he retires at age 65. In this ease education is considered a pureinvestment good and the marginal earnings profile is assumed flat. For

white men in this example, C=$5920; Et=$156; and t=49. For black men,

the opportunity cost is only $4720.

6This conclusion is fully consistent with other findings including

those of Bennett Harrison, Education, Training, and the Urban Ghetto

(Baltimore: The John Hopkins University Press, 1972) and Wachtel and

Betsey, op. cit.

157

140

(20.3%) more than the unaffiliated worker. In percentage as well as

in dollar terms, union membership is more helpful for the black worker

than the white. Stratification by industry also affects wages

significantly, although it is not as important a factor for black men.

A +l6 difference in minority employment ( %MININD) is responsible for a

$.24 difference in earnings.

The concentration ratio or "market power factor" also affects

earnings suggesting the existence of "complex crowding." A forty point

difference in the MPF (e.g. .20 vs. .60) is related to a $.36

difference in earnings. In addition, the government demand variable

is significant. Ceteris paribus, a ±10 difference in government

purchases means a $.16 wage differential. Apparently blacks do a little

better in industries subsidized by government contracts.

White Females - Schooling is the only significant human capital

variable in the white female equation; it yields approximately the

same wage increment as was found in the equations for white and black

men. This one factor is responsible for explaining about a fifth of

the variance in earnings.

The addition of the STRAT module increase the R2

to .511 and

reduces the SEE by almo6t twenty percent. Union membership and minority

employment by industry and occupation all affect white female earnings

after controlling for education. Union membership is valued at $.21

an hour yielding a percentage wage differential between union and

non-union workers approximately equal to that for white men. The ±Ia

evaluation of percent minority employment in the industry (MINIM)) and

s3

158

141

the percent minority in the occupation ( %MINOCC) yield $.30 and $.20

differentials respectively. Taken together these three variables

disclose a considerable degree of simple "crowding." if purely

additive, the three STRAT variables suggest a potential $.71 wage

differential around a mean wage of only $1 . As in the case of white

men, the industry module variables add nothing further to the explana-

tion of earnings for this segment of the labor force.

The nearly significant negative coefficient on "negative work

traits" can probably best be explained in terms of measurement error.

This is the only instance in which the coefficient is negative. In a

few cases the expected positive sign is found; in all others, with

this exception, the variable is insignificant.

Black Females - Once more the human capital module explains

about twenty percent of the variance in earnings. Schooling has about

the same dollar impact on the wage rate as it does for the other groups

in the occupation stratum. However, because of the extremely low mean

wage rate in this instance ($1.36), the rate of return on additional

schooling is greater than for any other race-sex group. In this case,

r > 4.25%. Training has no apparent impact on earnings although the

percentage of black females in this OCC STRATUM with training (7.3%)

is only slightly less than that for black men (9.5%). The other

significant variable in the module is migration. Remaining in the

same location after age 16 reduces the average wage by $.27 an hour,

similar to the effect seen for black men. In this stratum, migration

is an important human capital variable for blacks but not for whites,

159

142

most likely signifying the greater importance of emigration from the

south for nonwhite members of the labor force.

Both the stratification and industry modules add to the explained

variance in earnings boosting theR2

to .402 and reducing the SEE to

less than $.40. Union membership is worth $.28 an hour, somewhat

less than that for both groups of men but somewhat larger than the

impact of membership on the earnings of white women. Given the low

average wage rate for non-union black women in this stratum, unioniza-

tion increases the average wage almost 22 percent. This is approxi-

mately the same amount as for black men. The ±la evaluation of %MININD

results in a wage differential of $.30 an hour, identical to the

impact of industry segregation on white women and only slightly more

than the impact on black men. In addition, a ±la difference in after

tax profit rates is valued at $.18 an hour, an indication that

differences in industry demand also affects individual earnings.

For all four race-sex groups then, the stratification variables

are significant in this occupation stratum and have coefficients

of substantial magnitude after controlling for human capital. We take

this to be evidence of significant "crowding." Further analysis will

be postponed to Chapter VI.

Cross Race-Sex - Without resorting to Chow tests, it is evident

that there are some essential differences in the earnings generating

functions for the four individual race-sex groups in 0CC STRATUM 1-3.

While the same key variables are significant (schooling, union

membership, and %MININD), there are two important differences in the

160

143

regressions. In the first place migration plays a prominent role

in the functions for both black groups, but has no apparent impact

on the earnings of whites. The second difference relates to the

influence of the industry module; these variables are also significant

for the black equations, but not for the white. Differences in

industry structure apparently have no systematic effect on wage

differentials within each of the two white groups after controlling

for supply , ide stratification. On the other hand, the wage differ-

entials for both black groups reflect "complex crowding." Put

somewhat differently, differences in both supply and demand conditions

influence the earnings distribution for blacks while differences in

labor supply conditions alone appear to account for the explained wage

differentials of similarly qualified white workers.

This structural difference in the earnings functions appears to

be related to the relative variation in the underlying distribution of

industry characteristics. The significant coefficient on the government

demand variable in the black equation may be due to the fact that

the dispersion in this factor is much greater for blacks than any

other race-sex group. The coefficient of variation for black men

is 2.2866 while for white men only 1.6779. The same can be said for

the significant coefficient on after-tax profits for black women. Here

the coefficient of variation is .713 while it is no higher than .495

for any other group. Still again this holds for concentration in

the cross race-sex equation. The absence of significant coefficients

on the industry characteristics in the white equations thus may be due

161

144

to the fact that white workers are found in relatively homogeneous

industries while some blacks gain access to "permissive" economic

environments and others do not. Those who do enter the more concen-

trated, more profitable industries earn somewhat more than their

apparently misfortunate counterparts.

Although the cross race-sex equations mask these differences

in the structure of the earnings functions, they nevertheless con-

tribute to an understanding of wage determination by their ability to

estimate the impact of crowding even where segregation is near perfect.

On average across race-sex groups, schooling taken outside the

south contributes about $.07 to the wage rate per year of education,

although southern schooling is apparently worth less than one cent.

Vocational training adds $.39 to earnings while those who never migrate

from their place of residence at age 16 earn $.20 less per hour.

Altogether the human capita' variables can explain only 19 percent of

the variance in this stratum.

The addition of the remaining modules boosts the R2 to .52. The

$.38 wage increment due to Union affiliation is equivalent to an 18

percent differential between union and non-union workers, a figure

closely in correspondence with the early institutional results of

Levinson and similar to the more recent figures given by Lewis and

Stafford.7

At least for this occupation stratum, the early

7Levinson reported 14-18 percent in his early calculations;

Stafford 10-16 percent; Lewis 10-14 percent. See Harold Levinson,

"Unionism, Wage Trends, and Income Distribution, 1914-1947," Michigan

1sz

145

institutional results based on industry data were not badly biased

by the exclusion of human capital variables.

Both %MININD and %MINOCC are significant as well, together

contributing substantially to wage differentials. The ±1(7 evaluations

are worth $.44 and $.58 per hour respectively. If union membership

and the two minority employment variables are strictly additive,

market restriction induced crowding accounts for a measured wage

interval of $1.40 around a mean wage of $2.27. Differences in the

concentration ratio add another $.21 to the total measured wage differ-

ential. But how much of this is due to crowding and how much to

"pure" discrimination?

Equation (I) is the pooled regression with race and sex

variables added.

(I)

w = 3.1451 + .0452 Schoolirg -.0378 School-South + .3375 Training(3.45) (4.15) (2.37)

-.2061 Migration + .3664 Union Member -1.0019 %MININD(2.77) (4.62) (3.51)

-.7489 %MINOCC + .3887 Concentration - .1687 Physical Demands(3.27) (2.38) (2.41)

- .2439 Blackp - .3958 Fernald°(2.74) (3.86)

R2

= .558 Sr.: = .6042

Business Studies, Vol. X, No. 1 (Ann Arbor: Bureau of Business Research,Graduate School of Business, University of Michigan, 1951); FrankStafford, op. cit.; and H. Gregg Lewis, Unionism and Relative Wagesin the United States (Chicago: University of Chicago Press, 1963).

163

146

The coefficients on %MININD and %MINOCC both decline after

adding the race and sex dummies, but the fall is not especially

precipitous. The largest decline is from -1.1527 to -.7489 for the

coefficient on %MINOCC, but even this reduction is not particularly

significant. The t-statistic for a difference in the two coefficients

using the test of means is only 1.27, well below the level necessary

for a clear indication of statistical difference.

TABLE 5.2

REGRESSION COEFFICIENTS ON THE STRATIFICATIONVARIABLES IN THE POOLED OCC STRATUM

1-3 REGRESSIONS

UNION %MININD %MINOCC

Without R,S Dummies .3769 -1.1829 -1.1527(t value') (4.62) (4.03) (5.28)

With R,S Dummies .3664 -1.0019 - .7489

(t values) (4.62) (3.51) (3.27)

Reduction due toR,S Dummies .0105 .1810 .4038

2.8% 15.3% 35.0%

Union membership, %MININD, and %MINOCC are obviously not mere proxies

for race and sex, nor is the market power factor. Even after the

race and sex dummies are added, the total measured wage interval

for the stratification module is $1.12, eighty percent of its previous

value.

164

147

Together with the evidence from the individual race-sex

equations, the pooled regressions demonstrate that crowding, at least

within this one occupation stratum, is a conspicuous factor in

determining the distribution of earned income. In addition to "pure"

wage discrimination, industry crowding seems to perform an essential

function in determining wage rates for ach race-sex group, including

that of white men. Occupational segregation is also an important

factor particularly for white women. Finally unionization plays a

substantial role in the wage determination process, a finding con-

sistent with institutional analysis. Beyond these restrictions on

the supply side, the market power factor is significant suggesting

that demand side characteristics affect earnings, again in perfect

accord with traditional institutional theory.

OCCUPATION STRATUM 5

Occupation stratum 5 is .omposed mainly of semi-skilled manual

corkers. Almost two-thirds of the white men in this stratum are

sound in jobs under the single occupation title, "operatives and

kindred workers, n.e.c." Similarly 55 percent of the black men are

found in this occupation group with another 22.4 percent being

janitors and sextons. White women are less concentrated in the

operatives category; 38.8 percent are found here while another 14

percent are clerk typists, 12.7 percent are manufacturing checkers

and examiners, and 11.5 percent are assemblers. For black women,

45.4 percent are operatives. With the exception of typists, OCC

165

148

STRATUM 5 is the traditional semi-skilled blue-collar workforce.

The average full-time worker require. between one and three

months of specific vocational prepar-,,tion to perform his or her job

adequately. (SVP=2.9) The typical worker had a little more than nine

and a half years of formal schooling and almost 11 percent have parti-

cipated in some form of institutional training program. Average

experience in the labor force is thirty years. Eleven percent of this

occupation stratum is black while 39 percent is female.

The industries in which these individuals work appear in the

aggregate statistics to be similar to those in which workers in 0CC

STRATUM 1-3 are employed. A little more than half of the workfot..:e

in stratum 5 are union members while the average minority employment

in these industries was 35 percent. Each worker has slightly more

capital to work with: $24,000 vs. $20,000 in depreciable assets/

production worker in stratum 1-3. The historical average after-tax

profit rate is about .8 percentage points higher.

White Males - mile average wage rate for white males in this group

is $2.37, 16 cents higher than in occupation stratum 1-3. The "best

fit" human capital equation explains 16 percent of 11,0 variance in

earnings with schooling, school-south, and migration each contributing

to the regression. An additional year of school is valued at $.12

per hour except in the south where it returns two cents less.

Migration is worth $.18 an hour, migrants earning some 6.6 percent

more than those who have not moved since age 16.

166

149

The addition of the STRAT and Industry modules increase the R2

to .234. Unlike its positive effect on the other groups of workers

in this stratum, union membership does not appear to affect white

male earnings. Ceteris paribus, the two-fifths of white males in this

stratum who are not members of a trade union earn the same amount as

the 60 percent who are. Industry segregation, however, does have some

effect on relative earnings. The ±la evaluation of %IININD is valued

at 23 cents an hour. This is far less than in 0CC STRATUM 1-3, but

nevertheless still substantial. Concentration is the only significant

industry variable; after its addition to the equation no other industry

variables are significant at the .05 level. A similar ±10 evaluation

of concentration suggests a $.38 wage differential.

The positive coefficient on concentration in the face of an

insignificant union membership variable cannot be easily explained.

One possibility is that union membership is sufficiently colinear

with either concentration or %MININD that its real significance is

not registered in the regression.8

This hypothesis, however, is

belied by the fact that after the introduction of the human capital

module, the addition of union membership alone still does not yield a

coefficient which is significant at the .05 level. An alternative

explanation relies on the theory that relative wages are not correlated

with unionization because of the "spillover" effects or "sympathetic"

8The zero order correlation between union membership and concen-

tration is .229; between membership and percent minority employmentin an industry ( %MININD), -.244.

167

150

TABLE 5.3

REGRESSION EQUATIONS:OCCUPATION STRATUM 5 BY RACE AND SEX

White Male Black Male White Female Black Female Cross Race-Sex

Constant 1.8370 1.9263 2.1246 1.6893 1.3942 .7647 .0355 1.2803 1.8284 2.4466

HUNAN CAPITAL MODULE

Schooling .1221 .1033 .0733 .0478 .0678 .0484 .1362 .0982 .0816 .0600

(8.54) (7.33) (5.39) (4.47) (3.70) (2.97) (5.51) (4.31) (6.86) (5.61)

School-South -.0252 -.0254 -.0418 -.0242 -.0215 -.0149* --- --- -.0316 -.0251

(2.72) (2.85) (3.94) (2.92) (2.39) (1.87) --- (4.65) (4.18)

Training ---4 --- ---- -- --- .3058 .3118

--- (3.05) (3.53)

Migration -.1827 -.1540 -.3016 -.2120 --- --- -.1664 -.1349

(2.32) (2.05) (3.24) (2.91) --- --- --- (2.73) (2.54)

Experience --- --- --- .'-- .0177 .0121 - --

-- --- --- --- (3.21) (2.42) ---

Specific Voc. Prep.

STRATIFICATION MODULEUnion Member 0.00. .6240 .2525 .2720 .2625

(8.16) (3.36) (2.83) (4.68)

Z Minority--Industry -- -.8005 -.5723 -.5404 -1.1726

(2.91) (2.34) --- (1.96) (6.22)

Z Minority--Occupation0.00.0 -.4403 -.4564

- - (1.94) (2.57)

2 Black Male--Occupation - --

2 White Female -- Occupation - -- - - - -

2 Black Female--Occupation - -

INDUSTRY YODULEConcentration .6848 --- .8154 --- --- .5214 --- .4502

(4.74) --- (6.21) --- --- (2.65) --- (3.64)

Union x Conc.

After-tax Profit 000,40 10.2611 5.3120

(4.20) (2.54)

Capital/Labor Ratio .0104 .0019

(5.15) (3.80)

Government Demand 2.8019 -(2.26)

WORKING CONDITIONSMIOULKPhysical Demands --- -.2574 - --

(2.24) ---

Negative Work Traits

R2 .158 .234 .141 .495 .055 .292 .165 .378 .099 .333

SEE .8189 .7831 .7467 .5758 .7043 .6131 .6173 .5413 .8705 .7524

MEAN $2.87 $2.87 $2.39 $2.39 $2.01 $2.01 $1.86 $1.86 $1.48 $2.48

444 444 277 277 295 295 158 158 823 823

168

151

pressure of potential union organizing attempts. Non-union firms

may pay union scale to forestall organizing drives. In this case,

while unions may have an impact on absolute wage levels for all

workers, there is no discernible effect on relative inter-industry

rates. Concentration can still play a role in wage determination under

these circumstances. It measures the ability of an industry to meet

the prevailing standard set through collective bargaining in unionized

sectors of the economy.

Black Males - The structure of the regression equation for black

males in 0CC STRATUM 5 is similar to that of white males with the

exception of a significant and extremely powerful union membership

factor. The human capital equation explains 14 percent of the variance

in earnings with the same variables as in the white wale equation.

However, the effect of schooling on earnings is significantly lower

for black men. An additional year of education increments- the average

wage by only 7.3 cents compared with over 12 cents for white men.

This is a significant difference at better than the .02 level according

to the standard test for a difference in means.9 Using the internal

rate of return method presented previously, the return for white males

is approximately 3.5 percent while that of black men is less than 2.0.

As expected, migration pays off somewhat more handsomely for

black men than for whites in the same occupation stratum. Again this

is taken to reflect the importance of migration from the south.

9t' = 2.47.

169

152

Within 0CC STRATUM 5, membership in a trade union is critically

important in the earnings function for black men. On average, black

workers of this skill level who do not have access to a union earn

$.62 less an hour. Union members thus earn 30.5 percent more than

non-union workers, a percentage much larger than most institutional

estimates with the exception of those reported in the research of

Johnson and Youmans.10 Semi-skilled manual black workers apparently

are found in two kinds of industries: relatively high wage unionized

industries where they are paid wages not far below that of their white

male counterparts and relatively low wage non-union industries where

they comprise a disproportionate share of the workforce. Which industry

sector an individual can enter is crucial in determining his income.

The minority employment factor also helps to explain some of

the variance in earnings. The ±lci evaluation of 7.MININD is valued at

$.18 an hour. Concentration influences the wage rate as well. In

this instance, the ±10 evaluation results in a hefty $.48 earnings

differential for similarly qualified workers. The combined addition

of the two stratification variables and the market power factor

increases the fromfrom .141 to .495 and reduces the standard error

of the estimate from $.75 to $.58. Quite clearly the stratification

theory explains a large part of the variance for this segment of the

labor force.

10See Johnson and Youmans, op. cit.

170

lb3

White Females lhe "best fit" human capital equation does not

...-,

ce..plain much of the variance in earnings (R :-,.0.:6) for white women

in this stratum. An additional year of schooling add9 significantly

less to average earnings than it does for white men and the internal

rate of return on additional schooling is no more than that for black

men (2.0%). Schooling taken in the south ib worth only 4.5 cents per

hour per year. None of the other human capital factors are

significant at all.

Union membership is the only variable that appears in the

stratification module. Membership adds $.25 to the wage rate over

non-union workers in this group, an addition of 13 percent. The

absence of 7.MININD and %Mili0CC may be due to colinearity with the

union membership variable, but this seems unlikely given the relatively

small zero order correlations bemeen those variables:

Union Membership

° "MININD -.146

ZINO0C -.099

Concentration was highly significant when regressed alone on

white female earnings, but it consistently tended to undermine the

into ;city of the human capital module. Thus it was deleted according

to the estimation procedure and three other variables were used as

"quasi"-instruments: after tax profits, the capital/labor ratio, and

government demand. Each of thesevariables measures some facet of

"ability to pay" with the ::10 evaluations yielding wage differentials

of $.34, $.42, and $.18 respectively. Whether these effects are

171

154

strictly additive was not tested. After the introduction of the STRAT

and IND modules, the coefficient of determination rose to .292 and

the SEE declined by $.09. Once again the institutional hypotheses

appear to be valid after controlling for human capital.

Black Females - Unlike the white female results, both schooling

and experience are important variables in explaining the wage distri-

bution for black women in this stratum. Together these two explain

17 percent of the variance in earnings. Schooling is particularly

powerful adding 13.6 cents an hour to the wage rate per year of

education. This translates into a rate of return of 7 percent, much

higher than for any of the other race-sex groups. For some unexplained

reason, southern schooling does not detract from this return. Every

additional year of labor force experience also appears to augment

earnings, in this case by 1.8 cents per hour.

The stratification module is powerful as well. Both %MININD

and %MINOCC are significant factors as well as union membership.

Unionization adds about the same amount to earnings as it does for

white women, $.27. In addition, the ±la evaluations of %MININD and

%MINOCC are valued at $.22 and $.18 respectively.

As in The equations for white and black men, concentration is

also significant indicatilg a substantial degree of "complex crowding."

The ±lo evaluation of the market power factor indicates a $.30 wage

differential. Altogether, evaluation of the stratification and industry

variables suggests a $.97 wage interval around a mean of $1.86.

172

155

The physical demands variable is significant in this equation as

well, but once again its coefficient is of the "wrong" expected sign.

The negative sign may be explained by the possibility of black female

"entrapment" as janitresses. This is a particularly low wage job

which has relatively heavy physical demands although the educational

and training requirements are not especially lower than for operatives

or assemblers.

When all of the variables are added, the complete equation

explains more than twice the variance explained by the human capital

module alone. Thus even here where the human capital variables are

relatively powerful, stratification factors still play a significant

role in wage determination.

Cross Race-Sex - The human capital equation for the pooled

regression explains only 10 percent of the variance in earnings within

0CC STRATUM 5. In this regression additional years of schooling are

worth $.08 per year except in the south where they return only $.05.

Migration is also significant reflecting the importance of this

variable for both groups of men as suggested in Chapter III. In

addition, however, the training variable turns out to be powerful

($.31) and significant at more than the .01 level. Training was never

significant within the individual race-sex equations thus suggesting

the possibility that training has an effect on wages between races or

sexes but not within them.

There is a bit of evidence in the data that training opportunities

are greater for men than for women, at least in this occupation stratum.

173

156

Thirteen percent of white males in this group had some institutional

training and 12 percent of black men. However, only 7 percent of white

females and an insignificant number of black women were exposed to

vocational training. It is then possible that differences in training

within the white male group, for instance, are not important enough

to be manifest in a significant coefficient. However the difference

in training opportunities between white men and black women, for

example, may be great enough to generate the large positive coefficient

found on this variable. This conclusion is enhanced by the fact that

once the race and sex variables are added into the complete equation,

the coefficient on training falls from .3118 to .1898.11

All of this

may be taken as evidence that the "structure" of human capital endowments

is important in addition to its absolute "quantity."

Inclusion of the stratification and industry modules more than

triples the R2 and reduces the SEE by $.12. Union membership is worth

$.26 an hour while the two minority employment variables are valued

at $.41 and $.16 according to our standard ±10 evaluation. The

impact of industry segregation is thus nearly identical in both this

stratum and the lower skilled 1-3 group. This is not true for occu-

pational segregation. In the former stratum the standard evaluation

of %MINOCC furnished a $.58 wage differential. This should come as no

surprise, however. Occupation stratum 1-3 includes a broad range of

specific jobs while group 5 is overwhelmingly composed of industry

11An identical phenomenon will be fcund in the "grand" pooled

regression reported later in this chapter.

174

157

operatives. In this case we would expect to find the major differences

in earnings related to industry attachment rather than occupational

category. Occupational crowding plays a role in wage determination

even here, but apparently a minor one.

Concentration, after-tax profits, and the capital/labor ratio

are also significant in the pooled regression. Summed together the

three are worth a substantial $.64 an hour based on the standard

evaluation.

Adding race and sex dummies to the complete equation seriously

affects the coefficients on the STRAT and IND variables as well as the

value of the training parameter. Equation (II) reports these results.

w = 1.8742 + .0613 Schooling - .0205 School-South + .1898 Training(6.13) (3.60) (2.34)

- .1494 Migration + .4012 Union Member - .4138 MININD(3.03) (4.43) (2.13)

- .1944 MINOCC + .7027 Concentration - .4754 Union x Conc.(1.07) (4.56) (3.63)

+ 7.0780 After Tax Profit + .0016 Capital/Labor Ratio(3.20) (3.20)

+ 1.5487 Government Demand - .1620 Physical Demands(2.71) (2.57)

- .3246 Black - .6704 Female R2

= .433 SEE = .6955(4.09) (11.04)

After the inclusion of race and sex, three more variables become

significant while %MINOCC drops out of the equation. For one, the

union-concentration interaction term is now significant. Evaluating

175

158

this regression for union and non-union workers and at .20 and .60

concentration ratios (as was done by Weiss) elicits the impact of

unionization under competitive vs. oligopolistic conditions. It also

demonstrates the impact of the market power factor in a unionized

industry vs. an industry not covered by collective bargaining. Among

generally competitive industries with workers in 0CC STRATUM 5,

union affiliation increases the average wage by $.31 an hour or 13.7

percent according to evaluation of the regression equation at different

levels of concentration and unionization. In the more concentrated

industries union membership is capable of increasing earnings by $.11

an hour or 4.6 percent. Alternatively, greater concentration (from

.20 to .60) raises earnings by about 12.6 percent

Concentration (MPF)

20% 60%

in non-union

No union $2.23 $2.52 +12.6%

Union $2.54 $2.63 + 3.6%

+13.7% +4.6% +17.9%

industries and by 3.6 percent when a union is present. These results

are more consistent with those of Stafford than Weiss in that both

unionization and concentration are still significant after controlling

for the human capital variables.12 Workers who end up in concentrated

12Recall that in Weiss's study, the addition of personalcharacteristics to the regression all but destroyed the significance

of the concentration term. The statistically significant coefficients

176

159

unionized industries earn approximately 18 percent more than

nonunion labor in the competitive sector.

In addition to the no's significant interaction term, government

demand also affects wage determination in this stratum. A +la

evaluation of its coefficient elicits an additional $.14 difference

in earnings. Physical demands is the third newly significant variable

after the addition of race and sex. Its coefficient is negative,

possibly displaying once again the "entrapment" of black females in

low wage physically demanding jobs.

The coefficient on the race dummy is -$.32 while that on sex

is -$.67. After the inclusion of both variables the coefficient on

%MININD declines precipitously from -1.1728 to -.4138 and %MINOCC

becomes totally insignificant. Clearly only a portion of the wage

differential between race-sex groups in this stratum can be positively

identified as directly linked to industrial and occupational crowding.

Much of the differential may be due to either pure wage discrimi-

nation within specific industries or occupations or due to segrega-

tion between firms rather than between industries. The high degree

of colinearity between the dummies (i.e. sex) and the minority employ-

ment variables makes it impossible to definitively differentiate

these effects. (See Table 5.4.)

in our results can be explained by the improvement in the measurementof concentration (through the use of the "market power factor") andthe micro measuranent of union membership. See Leonard Weiss,op. cit., p. 108.

177

160

TABLE 5.4

PARTIAL (XtX) MATRIX FOR OCCUPATION STRATUM 5

Race Sex %MININD %MINOCC

Race

Sex

%MININD

%MINOCC

1.000 -.062

1.000

-.039

.451

1.000

-.060

.420

.302

1.000

Nevertheless the consistent appearance of the STRAT variables

in the individual race-sex equations suggests an extensive degree of

crowding and together with the race-sex variables demonstrates an

even larger degree of overall stratification. Within the full labor

force (at least within stratum 5) there are large wage differentials

tied to factors which measure racial and sexual discrimination--in

one form or another--after controlling for differences in human

capital endowments. In addition there is strong evidence that sub-

stantial imperfections exist within this stratum's labor market even

for white males. In this sense the traditional institutionalist and

social stratification arguments are strongly upheld by both the

individual race-sex equations and in the pooled regressions.

OCCUPATION STRATUM 6-9

Occupation stratum 6-9 is composed of a broad range of specific

occupations which demonstrate a definite distinction between "men's"

178

161

and "women's" work and "white" and "black" work in the American

economy. This particular stratum is also noted for being the most

heterogeneously &lined of the five occupation groups used in this

analysis. Each of its specific occupations falls within a narrow

range of required "general educational development" (GED) but

potentially spans a wide range of "specific vocational preparation"

(SVP) requirements. On-the-job training can range from just six

months to, in rare cases, almost ten years (see Appendix A). For

this reason many of the results are not comparable to those found

5n other more narrowly defined strata. In the whole spectrum of

occupations from least to most skilled, we will find the regression

results for this group to 13 the most anomalous.

Over 46 percent of white men in this stratum are found in just

four specific occupations: truck and tractor drivers, general (semi-

skilled) carpenters, welders, and policemen. The four most popular

occupations for black men include truck drivers, but the other three

are shipping and receiving cletlrs, seek clerks, and hospital attendants.

Almost 55 percent of all black men in this stratum are found in these

occupations.13

13This comparison probably understates the difference in occupa-

tion categories for white and black men. There is no distinctionbetween long-haul and intra-city trucking in the specific occupationcategories given by the census. If such data were available it wouldprobably indicate that white men dominate inter-city trucking whilemost black truck drivers are found on local routes. Earnings are

considerably different for the two kinds of work.

179

162

White women are found in a different set of occupations

altogether. Almost 60 percent are found in just three occupations:

sewers and stitchers, hospital attendants, :nd receptionists. More

than 74 percent of black women are consigned to jobs as hospital

attendants, practical nurses, and sewers and stitchers in the apparel

industry. This extreme occupational segregation is one of the main

determinants of wage differentials according to the regression analysis.

The average worker in OCC STRATUM 6-9 had no more formal

education than the typical worker in OCC STRATUM 5, nor any longer

labor force experience. However the specific on-the-job training

required for these occupations is somewhat greater, as we noted, taking

in most cases between six months and a year to complete and in a few

cases more. Almost 13 percent of the workforce reported enrollment

at some time in an institutional training program. Only 9 percent of

the workers in this stratum are black and only 27 percent ace female.

Again over half of the stratum's workforce are members of trade

unions, but the variance by race-sex group is extreme.14

Fifty-six

14The large variance in union membership by race and sex is

found not only in occupation stratum 6-9 but in all other strata aswell. White men are more organized than black men and both male groupsalways exceed the u:;21,,nization rates for both groups of women. The

mean union membership rates by stratum are reproduced belOw. (See

Appendix B.)Union Membership Rates (%)

Occupation Stratum 1-3 5 6-9 12-14 15-17

White Males 58% 60% 56% 44% 11%

Black Males 47 55 44 37 n.a.

White Females 32 41 43 10 10

Black Females 29 47 40 n.a. n.a.

Total Workforce 50% 52% 52% 38% 11%

180

163

percent of white men are union members; but only 44 percent of black

men; 43 percent of white women and 40 percent of black females.

The average occupation has a minority workforce of approximately 30

percent, but as expected the standard deviation for %MINOCC is as

large as the mean.15 What is ironic about OCC STRAT 6-9 is that the

segregation of the workforce appears to be so complete that the strati-

fication and industry variables are not particularly important within

individual race-sex regressions. In this case, only the pooled

regressions can uncover the effect of "crowding" due to racial and

sexual segregation. Table 5.5 contains the regressions for this

stratum.

White Males - Both the human capital equation and the complete

equation for white males have few significant variables. Only

schooling is important in the HC module and this one variable explains

only 6.3 percent of the variance. An additional year of education is

worth a relatively small $.09 an hour.

The important factor in this stratum is union membership.

Consistent with what is generally known about the specific occupations

in this stratum, unionization is worth more than $.76 an hour thus

forging a 28 percent wage differential between union and non-union

workers. In no other occupation group is union membership so important

for white men. In addition, there is a significant coefficient on the

15The actual coefficient of variation on %MINOCC is 1.08 while

that on %MININD is .63.

1.81

164

TABLE 5.5

REGRESSION EQUATIONS:

OCCUPJ.TION STRATUM 6-9 BY RACE AND SEX

;Mite Male Black Male White Female Black Female Cross Race-Sex

Constant 2.0699

HUMAN CAPITAL MODULE.0927

(4.30)

001000.

00000040

------

000000

Oa Os /0

.063

.9506

$2.96

279

1.9847

.0791(4.04)

------

---

---

.7647

(7.28)

-2.5825(2.31)

.238

.8601

$2.96

279.

2.2122

.07640.45)

-.0477(3.27)

00

-.5384

(4.18)

4004E010

------

011.0.100

.011.00.0

00 amp

------

Ow+ Oa

40 .10.00

.156

.8725

$2.36

186

1.6901

.0597

(2.96)

-.0275(2.02)

-.3526(2.93)

.5961(4.42)

41.1

0000.=

.6535(2.77)

40.01

.320

.7874

$2.36

186

1.9361

------

--

-.1917

(2.08)

00d000

-----

.0.000

.1110.01.0

110.001.

400.10.

40.00

OPOD

010.1.

- --

--

001

0000010

.040

.4733

$1.84

106

1.6422

---

-----01.M00

-.1014*(1.12)

do.*

.3328

(3.54)

00

410

.6042(2.57)

- -

4=00 00

.166

.4453

$1.84

106

1.2439

.0678

(2.29)

-.0333(2.50)

----

00011.00

00004=00

110.1000

-----

----

004040

-----

.187

.4416

$1.72

43

1.5608

.0506(2.23).

-.0317(3.18)

,10100

MIME.

00.01.

.4966(4.57)

- - -4104,00/,

-4.4168(2.86)

0001.00

000000

01.040

-6.0458(3.52)

---

.620

.3140

$1.72

43

1.2216

.0560(3.15)

-.0318(3.05)

00

.2021

(4.62)

0.0 00

------

------

---

.04000.

---

01.004.

0040.0.0

---

- --

.091

.9563

$2.60

423

2.6503

.0511

(3.45)

-.0196(2.18)

.000000.

=0110

-- -

-

.6746(8.21)

-1.2192(4.37)

-.7525(3.73)

-z.4924(2.67)

40,0/4/0

00.0000

.00.06

0000

.380

.7927

$2.60

423

Schooling

School-South

Training

Migration

Experience

Specific Voc. Prep.

STRATIFICATION MODULEUnion Member

X Minority-Industry

X Minority-Occupation

Black Male-Occupation

X White Female-Occupation

X Black Female -- Occupation

INDUSTRY MODULEConcentration

Union x Conc.

After-tax Profit

Capital/Labor Ratio

Government Demand

WORKING CONDITIONSMODUIFPhysical Demands

Negative Work Traits

R2

SEE

MEAN

N

182

AMlb..--.011111=10.-.011.

165

variable for percent black male employment in an occupation (%BMOCC).

The usual ±lo evaluation of this factor yields an added wage

differential of $.24 an hour. Together these two stratification

factors are thus valued at $1.00 or just slightly less than a third

of the average wage. After their addition, the industry module adds

nothing suggesting that differences in demand characteristics mean

relatively little if anything after supply constraints play their role.

While not an especially well-specified equation, the regression indicates

that (1) differences in human capital are not particularly important

in explaining wage differentials within this stratum and (2) that

labor market stratification is the main actor in determining the

distribution of earnings.

Black Males - The black male equation is similar to that of white

men with the exception of the. importance of migration and concentration.

An additional year of schooling is worth 7.6 cents an hour which

according to the means test is not significantly different from the

schooling coefficient for white men in this stratum.16

Again however,

as in 0CC STRATUM 1-3, the rates of return on additional schooling

are so low for both groups that this apparent equality is not

especially valuable for black men. Schooling in the south is worth

even less, yielding only $.03 an hour.

As in every other black male regression, migration is highly

significant and powerful. Those who do not migrate during their

16t' = .57.

183

166

lifetime earn $.54 less per hour. For a full-time full-year worker

this is equivalent to almost $1100 a year.

As noted previously, industry and occupation segregation by

race and sex can be so extensive that %MININD and %MINOCC would not

be significant variables within !ndividual equations. This is

apparently true for the black male regression. Neither of these

variables is significant. However, union membership is, and once

again as in OCC STRATUM 5 its effect is robust. There is a $.60

wage gap between union and non-union workers which represents a 28

percent differential exactly the same as for white men. This is

roughly equivalent to the difference in wages found between a black

unionized maintenance painter ($2.75) and a skilled non-union hospital

attendant ($2.15).

After the inclusion of union membership, concentration adds to

the wage differential as well suggesting a "ase of complex crowding.

Unlike for white men, industry demand characteristics apparently

affect relative wages. The ±la evaluation is worth 35 cents an hour.

With unicnization and concentration included in the regression, the R2

more than doubles to .320 and the SEE declines by $.085.

WhiLe Females - The human capital equation for white women in

this group explains only four percent of the variance and neither

education, training, experience nor specific vocational preparation is

significant in this regression. Only migration is significant and

its coefficient is a relatively small -$.19. The addition of union

2membership and concentration raises the R to .166, although this occurs

184

167

only after the estimation rule concerning the integrity of the human

capital module is violated. Union membership is worth $.33 an hour

or 19.7 percent. The ±la evaluation of the concentration ratio is

valued at 23 cents an hour. Although this regression is not

particularly well fitted, it suggests that within this stratum almost

all of the explained variance in earnings is due to industrial and

occupational attachment. This is not an unreasonable conclusion given

the heterogeneous set of "women's" occupations represented in STRATUM

6-9, all of which have very similar GED and SVP ratings but differ

in terms of industrial characteristics.

Black Females - In contrast to the white female equation, the

regression for black women is quite servicable. The human capital

module explains almost 19 percent of the variance with the two variables

schooling and snocl-south. An additional year of education is worth

6.8 cents excepting the south where its value is less by 3.3 cents.

The addition of the STRAT and IND modules increases the toto

.620, the highest of any equation in the analysis. The standard error

of the estimate is only $.31 after the introduction of these modules.

Union membership is worth almost $.50 an hour. Th's represents a 33

percent differential between union and non-union s,orkers. A further

indication of the effects of labor market crowding is found in the

±la evaluation of percent black female employment In as occupation

(%BFOCC). This yields an additional differential of $.30. For

occupations that are even more "impacted" with black females the

effect is, of course, larger. The wage rate for hospital attendants,

t85

168

for example, is $.57 an hour lower than in occupations with only the

average percentage of black females. Quite obviously occupational

"crowding" severely affects earnings in this stratum.

In the industry module there is an unexpected sign on the

government demand variable. Here we find that a tla increase in

government demand apparently lowers the wage rate of black women by

nearly $.19 an hour. This is the only instance in the entire analysis

when a statistically significant government demand variable had a

negative coefficient. This rather puzzling result was found to be

merely a function of the peculiar industry composition in the micro

data for this regression.17

No other industry variables were signifi

cant so that our overall assessment is one of "simple"--but extensive--

crowding.

17Since we do not have a measure of government demand for the

"hospital" industry in our data set, the coefficient cannot be dueto the lower wages paid to hospital attendants in government subsidizedhospitals. Another solution was necessary. The puzzle was finallysolved after perusing the original data on government purchases byindustry. The weighted mean level of government expenditures in theindustries in this regression was 2.8 percent of gross sales. Thereis one industry in the whole data set which sells a larger percentageof its product to the government and also emp.,ys a relatively largenumber of black women in this o:cupation stratum. This industry is"Miscellaneous Fabricated Textile Products" employing a large numberof sewers and stitchers. It sells over 4.5 percent of its annualproduction to government agencies. With a low market power factor(.1043), a relatively low profit rate (2.5%), and an extremely highproportion of minority workers (60.7%), the average wage rate in theindustry is relatively low ($2.10 an hour in 1967). Black women,according to the data, have little access to other industries whichhave large government contracts, but also higher average wages.Consequently, in this single case, there is a negative relationshipbetween government purchases and individual earnings. This result isnot inconsistent with the general stratification theory.

186

169

Cross Race-Sex - Because of the extensive occupational segre-

gation in this stratum by both race and sex, there is a strong tendency

for the individual equations to underestimate the impact of crowding

on the distribution of earnings. Consequently the pooled regressions

may give better estimates of its effect.

In these equations the human capital variables are responsible

for explaining only 9.1 percent of the variance in earnings. Formal

schooling has a relatively weak impact on wages, an additional year

addino only 5.6 cents to hourly earnings and less than half as much

if the schooling was completed in the south. However, for the first

time in the analysis, specific vocational preparation is significant

suggesting the critical nature of on-the-job training in the occupations

within this group. The fact that SVP is statistically significant

in the pooled regression while insignificant in each of the individual

equations suggests that access to jobs wilich provide apprenticeship

(or other forms of investment in OJT) is linked directly to race and

sex. White males have freest access to training while the other race-

sex groups are provided with a lower average level of SVP.18

18The mean SVP scores for each race-sex group are:

SVP asvp

White Males 4.79 1.15

Black Females 4.37 .82

White Females 4.35 .78

Black Males 4.33 .93

Crcss Race-Sex 4.65 1.07

IS?

170

The addition of the STRAT variables compromises the integrity of

the human capital module in that SVP now disappears from the

regression. In this case we have permitted this to occur for while

the letter of the estimation procedure is violated, we feel its

"spirit" is not. The original intent of using a surrogate "two-stage"

regression was to account for the sequential relationship between an

individual's acquisition of human capital and his or her subsequent

attachment to a specific industry and occupation. In the special case

of SVP the presumed causal ordering is often the reverse. On-the-

job training such as apprenticeship is only available after a worker

has gained access to a particular job or occupation. If access is

denied on the basis of race or sex, as it often appears to be in this

group of occupations, then SVP acts more as a proxy for stratification

than a traditional human capital factor. Not to violate the original

estimation procedure under these circumstances would lead to

seriously downwali biased estimates of the effects of industrial and

occupational segregation.

Before the race and sex factors are added to the complete

equation, the introduction of the stratification module boosts the R2

to .38 and lowers the SEE by over 16 cents. Union membership alone

is worth $.67 an hour; those in organized industries or occupations

consequently earn a full 30 percent more than those who are not. In

addition, %MININD, %MINOCC, and %BMOCC are all highly significant.

If strictly additive, the sum of the ±la evaluations is worth over

$1.30 an hour, half the mean wage rate. Although quite hefty, .his

188

171

result is by no means unreasonable. It constitutes a differential not

much larger than the actual differences in mean earnings between white

men and white and black women. After the stratification module

variables are added, none of the industry variables are significant.

"Simple" crowding appears here to be the rule.

As expected the introduction of the race and sex factors severely

reduces the strength of the variables in the stratification module

with the exception of unionization. With the extreme occupation

segregation found in this stratum, the complete pooled regression

attributes much of the variance in earnings due to what we feel is

occupational attachment to the micro measured race and sex dummies.19

Equation (III) is the "best fit" complete regression including these

variables.

(III)

w = 1.8476 + .0532 Schooling + .0799 SU) + .6963 Union MilLbership(3.74) (2.23) (9.22)

- .7200 %Mi:tIND - .5476 Black - .6763 Female(2.90) (4.18) (5.41)

2R = .421 SEE = .7659

Union membership continues to be worth almost $.70 an hour clearly

labeling access to a un_o%ized occupation as the surest admission

ticket to higher earnings in this stratum, afT controlling for human

19The zero order correlations for %MINOCC and sex = .810

%BMOCC and race = .560

189

172

capital. SVP is significant once again, but use of the difference

in means test indicates that its coefficient in the final equation is

statistically lower than in the equation which contains only the human

capital variables. (C=2.16,

All of these results are consistent with what we know about the

specific occupations within this 0CC STRATUM. The Teamsters' and the

car: enters' union, for instance, have historically won wage packages

which are among the highest within the occupation spectrum. Workers

who gain access to these jobs or the other crafts gain from the

collective bargaining efforts of their unions; those who for one

reason or another do not enter these occupations will normally earn much

lower wages, ceteris paribus. This particularly affects the earnings

of women, but also limits the earning power of black men. The effect

of stratification, either in the form of "crowding" or -ure discrimi-

nation, appears to reach its peak in this stratum. Again, however,

because of the high degree of segrJgation, it is statistically

impossible to differentiate Jetween the two forms. Differences in

human capital--with the exception of SVP--only account for a small

fraction of the explained variance in earnings.

OCCUPATION STRATUM 12-14

For men, 0CC STRATUM 12-14 primarily contains individuals who

are defined by the Census as "craftsmen, foremen, and kindred workers."

The largest specific occupation for white men if "foremen" (17.7%)

followed by "mechanics and repairmen, n.e.c." (14.5%) and "machinists"

190

173

(8.5%). "Linemen and servicemen," "plumbers and pipefitters," and

"electricians" are also members of this group. For black men, the

largest single group is "mechanics and repairmen, n.e.c." (19.5%)

followed by "auto mechanics" (14.5%) and "foremen" (12.1%).

In contrast to both groups of men, most white women in this

occupation group are found in white collar jobs; almost two-thirds

(64.7%) are classified as "secretaries." Much smaller percentages are

found as medical and dental technicians and department heads and

buyers in retail outlets. The number of black women in this stratum

is so small that no statistically significant results could be obtained.

The typical worker in this category had eleven years of schooling

(10.8 years) and over 30 percent reported receiving some form of

institutional training. White females had, on average, more schooling

(11.4 years) but only 21 percent reported previous enrollment in a

vocational education program. The average occupation in this stratum

requires between two and four years of on-the-job training according

to the SVP scale. A few jobs require more.

Consistent with the specific occupational composition of this

stratum, a much smaller percentage of workers are trade union members

(38.4%). Foremen have been discouraged from organizing into unions

since Taft-Hartley and few secretaries and mechanics are members.

Only about 4 percent of the stratum is black and only 15 percent are

women reflecting the dominance ,f white men in these higher level

occupations. Table 5.6 contains the regression results for this

stratum.

191

4

174

TABLE 5.6

REGRESSION EQUATTONS:

OCCUPATION STRATUM 12-14 BY RACE AND SEX

White Male Black Male White Female Black Female Cross Race-Sex

Constant -.2914 .2250 1.9477 1.9279 1.0390 -.0761 --- -2.9893 -.7516

RUYAN CAPITAL MODULESchooling .1593 .1244 .0719 .0545 .1152 .1068 SAMPLE SIZE 4560 .1409

(8.35) (6.51) (2.76) (2.34) (2.66) (2.52) (8.91) (8.49)

School-South -.0397 -.0385TOO SMALL

-.0407 -.0370(4.58) (4.64) --- MINEN ---

FOR.(5.15) (5.00)

Training --- .5840 .4125 ---

--- (2.65) (2.14) --- --- SIGNIFICANT

Migration -.3327(4.08)

-.7991(3.84)

-.3493(2.03)

-.3126

(2.10)

---

---RESULTS

-.3148(4.20)

-.2515(3.63)

Experience .0097 .0070 --- .0166 .0101 .0084(2.55) (1.89) --- (2.16) --- (2.81) (2.55)

Specific Voc. Prep. .2893 .2606 --- . ...- .6594 .2982(2.45) (2.10) --- (8.08) (2.60)

STRATIFICATION MODULEUnion Member -- .3813 - .4705 -- 0 .3302

-- (2.58) --- (2.97) --- NOM 11111.111.M -- (2.40)

X Hinority--Industry --- -.7850 --- -.8673 --- .1 --- -1.0525--- (3.18) --- (2.14) (5.25)

Z Minority -- Occupation .4151111 , (2.16)

2 Black Male -- Occupation --- -8.0053 ---

-- (3./q) - -Z White Female-Occupation --- 1111.1111. -- -Z Black Female-Occupation --- MOWN* -INDUSTRY MODULEConcentration -- .9911 --- .8951 --- --- 1.1334

--- (5.56) -- (3.24) --- (7.02)

Union x Conc. -- -.7066 - -- -.7677

--- (2.77) --- --- (3.24)

After-tax 9rofit .0 --- 18.6527 - -- --- 6.3150- (4.61) --- --- (3.05)

Capital/Labor Ratio .0071 ---

(2.33) ---

Goiernment Demand 1 01.0

WORKING CONDITIONSMoDITEPhysical Demands WM* MI1111.

Negative Work Traits --- .1187 --- -- .1218--- (2.98) --- -- (3.23)

R2

.160 .252 .193 .424 .059 .277 --- .203 .336

SEE 1.0344 .9799 .9091 .7773 .8962 .7960 --- 1.0530 .9654

HEAN $3.50 $3.50 $2.59 $2.59 $2.36 $2.36 --- $3.29 $3.29

674 674 128 128 115 115 - - --- 820 820

192

a

175

White Males - All of the human capital variables with the

exception of insfitutional training are significant in the white male

equation. An additional year of education is worth almost $.16 an.

i

,

hour; in the south almost $.12. Non-migrants earn $-.33 less per hour

-.-----____ .

while each additional year of labor force experience is valued atd C q

almost $.01. Apparently experierde only begins to play a role in.1 -

earnings functions in the relatively skilled eccupations; within each

of the lower level,stratalthe wage-experience profile is flat. Finally

each unit in the SVP scale adds $.29 to,the hourly wage. Taken

_tegetherthese five factors explain_16 percent of the variance in

earnings.

Even within this relatively high skilled occupation stratum,

the stratification and industry_modules are reSpOnsible for a large._

f: -2 'increase in the coefficient of determination. The 4 rises to :252-

-after these variables are entered. The differential-in earnings due

o industry segregation -is almost $.25 an hour Which is eqdivalent1

--to_a 7.3 percent wage differential. There an Additional $.32

or 9.7 percent differential due to the crowdi4; of black men into

:-certain occupations = ( %BMOCC).

11 f ft

Union membership and concentration each increase the wage ratek

_

as well, but again the interaction term is negati Among. generally

k

competitive industries (MPF-4.20),_union membership adds $.'24 to-Y

hourly earnings or-7,4\percent. 'Among concentrated industries, however,

union membership apparently adds nothing extra to the wage rate.

_ Above a concentration ratio of .54, in fact, the statistical

193

,No_ union-

Union

176

Concentration (MPF)_

-20%'

$3:25 -0.65

$3.49 $3.60

-relationah*between unionization and_earningsis Slightly, negative,"

,In:the unorganized sector oligopolies pay an average of 12.3 percent

more than do competitive industries. In 'the union sector a 40 point

differential in the market pOwer factor is worth no more than an

additional 1L-cents or 3.2 percent. Yet, overall,. after we control fdr

human capital differences,ta unionized highly concenttated industry

'N\

pays wages which are $.35 an hour or 10.8 percent greater than an

industry which is unorganized and competitive. The importance of the

industry and stratification_variables thus prevails even among

relatively highly-skilled white ten.

One other point might be added. In the complete equation; the

negative working traits factor in the working conditions module is

The apparent negative effeCt of union membership on wage'rates in highly concentrated industries may-, in fact, have somesubstance to .it, It seems plausible that non -union concentrated=industries may pay higher straight-time wageS tri order to ward_off the

,sympathetic pressure for unionization. The unionized induserieS.on the

other hand may-settle on a total economic and rion:-economic_package_-:which-realizes a lower straight -time rate, but makes-up for it with_-_larger fringe benefits including_ longer 'vacations,'more numerous-_holidays, fully-paid medical insurance, life insurance, large pen-sip-rig,:

etc. In this case, unionization may mean lower straighttime wa3exates

but higher total remuneration. The regression cannot measure this -_

effect.t

i 194

4'

177

significant and of the expected positive sign. The average occupation

in this stratum for white men has_approximately one (.93) negative

, .---

'trait with a-standard deviation ofnearly one (.98).,-After controlling..'

for all other factors, a worker in an occupation with one extra

.

negative trait earns $.12 more per hour in compensatory payments.

What this appears to signify is that. amonerelativeliskilled (white

male)- workers, but not among the unskilled, wage rates/respond to

the '!quality",of the job in a compendatory manner..

Mack Males - An nsadditional year of education is worth-signi,-

Idantly 4.es6 to a black-male Worker than to a white male_in

=_= =stratum, What is more*-the-rate_of return:on schooling is--the_

lowest_of anrccupation group for black men refle!cting both.the-

higher oppOrtunity costs of additional education and the near constant

= /2-dollar value_ of schooling exhibited in,,each=of the strata. On the-

,

r hand, institutional- training pays'off handgomely, contributing

8 an hour to the wage rate. This most likely represents the

urn to specific training in fields like auto mechanics. As usual

ration is an important`mportant Iactor for black men, contributing hereft

9

21The differencelin meads test yields a e=2.57.

22The-dollar values and the internal rates of return for a year -_

schooling for black men by occupation stratum are:

ccu ation Stratum Dollar Value Internal R=Of R -

: 4733 -.0764

.0719n. a.

195

2.0<2.0<1.25

178

_almost $.35 an hour. The value of migration is similar to that found

inOCC STRATUM 1-3 end 5 although below that in stratum 6-9.,

Even within thaffrelatively skilled grOup,.the additi9n of theA w .

z

2-STRAT and IND modules more than doubles-theit to .424 and lowers _the_

.._

SEE_by $.13. Union membership is Worth $.47 an hour and there is no

apparent negative interaction between unionization and concentration

as there,is far white men. This may be due to the fact that on

,average black workers-in this stratum are found in much lesp-concen-

- .... . .-_

.

_trated industries. The mean MPF for white_men is'.46; that_for black-

16en only -.36. The regtession,_in this case, was not capable.-of

iso.,,ating the impact of concentration on the wage effect-of union

_membership_for the few black,workers_in oligopolistic industries.Il

"-Translated into percentage terms, the $,.47_wage increment dtig_torndion- ,,, .

affiliati iog issresponsible for a 19.5 percent wage differential;

6

--,

rate more than twice as large as thatfor:white men in the same stratum.

A

_ __-

This adds to the mounting evidence that unionization while important

for_white men is much more'So for blackmales in every single stratum.__

'eln this caste, access to the organized skilled crafts is the port of

entry to_higher earnings, human capital constant.

Industry segregation has an additional effect on relative

earnings, The ±laevaluation,of =MIND is valued at $.32 an hour..

In dollar terms, this is a bit larger than for any of the other-strata._

Tn the industry module, the same evaluation of concentration is valued

a6 $.54, indicating that the few

oligopolistic industries benefit

1.96

black men who do gain access to

.3°

substantially. Here "complex"

I -179

crowding seems to prevail. If strictly a'dditive, union membership,

DfININD, and concentration would furnish a wage differential of,,

$1.33, more than half the mean- wage for this group: This is, , ,

,;

equivalent to wage rates of $1,93 vs. $3.26 an hour.

White Females - Before. the introduction of the industry module111.

the only significant human, capital variable is schooling. Each _year

is worth 11.5 cents an hour, higher than in any previous group. After-v.-

the addition of the 'industry Module years of experience in the labor

force also appears as a significant factor in the earning generating

'function. EaCh year is valued 'at about 1.7 cents per houi which

_

ranslates into a wage differential of $.33 between a ilortan yho is

years of age and -one whof is 45..

_None of the- factors in the stratification in odule are significant_= 0

_including union membership; yet two of the induStry or "demand" side

variables are. According to theory there_ must be skther factors

-beclide unionization and minority segregation which serve to segment

A

the labor force. Imperfections in job information and inter- and

intra-labor market immobility due to geographical barriers may be

factors sufficipnt to offset the tendency toward wage equalization

as indicated by statistically significant variables in the industryse

These results, are, understandable in light of this stratum's

occupation composition. With 65 percent of the workforie as

"secretaries, only 10.percent of the workforce unionized, and -the mean

percent of white females in an occupation greater than 70 percent, the

)

187

t..

n

_

180

°majority of the variance idearnings, after pontrol/Ing for human

capital, is pr'bably due*to the pure institutional factor of "ability_, ,

to pay." Within the stratum, *Parity crowding and unionization are-

not particularly important, but which industry a secretary has access4

to apparently affects tie wage. A secretary in-an industry which has

a profit rate tlo above-tile mean will earn, on average, $.35 more per

. _

hour: In an industry which has ±]x more capital per.production worker

the average wage will be $.1T higher.23

The addition of just these

two variables is sufficient to more than quadruple the explained

variance and again suggests the tremendous importance of industry

attaqhment tor minority members of the labor force.!

'Black Females - SAMPLE SIZE TOO SMALL FOR STATISTICALLY

_ 4

Cross Race -Sex As in occupation stratum 6,9,_the individual

race-sex equations may underestimate the impact of.crowding. This is

particularly true because of the nearly 4mpfete segregation of white

women_in this stratum: The pooled regressions are therefore of value

n attempting to identify the true relationship between "crowding"

=and earninge;=

With the exception ot_institutionaltraining, all of the fattOrs

--in --the human capital module in the Pooled regression are highly

23It also seems plausible that "quasi-"siMpathetic pressure may

work within an industry. White collar personnel in indostries withstrongly unionized blue-collar wOrkforces may.benefit from the pro-duCtion workers' union without belonging. In the present_egyatio_n =it

is possible that, profits and capital-intensity are correlateewith theextent of blue collar unionism and consequently, produce this phenomenon.

188

1514

significant and explain about a fifth of the variance in earnings.

Unlike the weak effect of schooling in the previous stratum, here a

year of education is worth in excess of $.15,an hour. (In th& south,

an additional year of school is valued at 11.5 cents.) Migration.

adds over .$.-31 to av4rage hourly earnings and each year of labor.

t'force experience adds another $.01 0to the wage rate.

As expected, specific vocational preparation (SVP) is a potent

factor in the earnings function reflecting the,prime importance of

apprenticeship in the skilled trades. ,At the same time, the absence

of this factor,in-the equations for black men and white women and its

-Weaker presence in the white equation_ exposes the nature of:the_

link between demographic_characteristics and access -o occupations-_2

wh'ch-offer apprenticeship. The link runs first from race_and'sex_

upation and then from oCcupation to specific training. Access to

-fjob with an SVP _rating one -unit=higherthan-themean is worth___-_,

neatly $.66 an hour,- The addition of the stratification,,industry;._

and working conditions modules redUces the coefficient on _SVP by more_

-ban half and the further addition of the dummy for sex eliminateS

SW altogether._ Once again we have allowed this to occur because o

the- presumed causal relationship involved in the function.

The final complete equation including-variables for race'and

a

sex is shown in Equation (IV).

199

45

182

w'= 1.7244 + .1264 Schooling - .0354 School-South - .2601 Migration(7.52) (4.89) (3.84)

+ .0088 Experience + .-3991 Union - .8550' %MININD(2.98} (4.29)(2.66)

0- 7.3783 %BMOCC + .9492 Concentration - .8240 Union -Conc.

(4.55) 0 (6.10) (3.60)'

. .

+ 5.5602 After-tax Profit + .1304 Negative Work TraitsX2.78) (1v59)_

- _.8806 Sex

(8.28),: R2 = .370 SEE -= .9411

Except for SVP, the ineirity of the original human capital

equation_ia preServed-.

The dummy variable for_Tace is insignificant after controlling

_JorIMININD and black_male employment __(%BMOCC).. _141hile there may -

_still be some pure racial_disorimination within_industries, occupation,)-- =

and_specific_

rAilated more

AMoreover the

_firms,_the dominant stratification effect appears to be

'directly industrial and occupational crowding.

crowding hypothesis is supported by, evidence that the

Inclusion of the sex variable has only a minor deteriorating effect

oh-the coefficient_on MININD.and'none on %BMOC.C. Without the dummy

?the-±lo evaluations of these two factors are wroth $.37

:7ands $.23 respectively.. After the dumMy is added, the value on %MINIMA

falls by tinly 7 cents and the coefficient On %BMOCC actually rises

by $.11. Thus pure sex discrimination exists simultaneously with

crowding leaving the average female,$.88 an,-..hour worse off.

20 04

183

j On the demand side, union membership and concentration interact

in an almost identical fashion in the pooled regressidn as in the white

male equation. Among unorganized workers, a forty point difference

in the concentration ratio is responsible for a 12 percent difference

in average hourly earnings. In Competitive industries, union member-

=-ship is worth about $.23 or 7.4 percent; in concentrateCindustries,

unionization adds nothing to the wage rate. All in all there exists_ -

a 9.0 percent wage differential between similarly skilled workers in

unionized concentrated and unorganized competitive - indus- tries.

-union

Union

= Concentration-_--OTYY:-

-.207

+7.4%

_ Higher after-tax,profits also affect earnings in this stratum-_

independent of_unionization and concentration. The average wage rate

in an industry with after-tax profits ±la_higher=than the mean is-

.10 greater than for workers in the "average" industtf.c As in

white male eqbatison, t gative work traits is also significant-and--

the-expected positive sign, indicating again that at least at this

higher skill level, compensatory wage payments are necessary indUCe

-a-sufficient supply of labor to the more "distasteful" jobs.

What we,find_most interesting about .these results is that they

show that even' among relatively well-educated and skilled workers,

201

184

large wage differentials can be traced to factor's other-than differ-

ences in human capital and working conditions. This begins to

disappear only among the ver y most educated and skilled workers; those

--- in occupation-Stratum

OCCUPATION STRATUM 15-17

Unfortunately the number of blacks in tke,SEO's highest skilled

occupation stratum is too small to allow individual statistical

analysis---1Consequently the retalts refer only.to the white population

1

_-except for the pooledrace-sex-equations.

- -_i--,i-managerial occupations within this stratum he largest numbers are

found as accountants, insurance agenta,_draft Men, and secondary---...

White men are found in a plethora of sp cific profI essional and

*school teachers. Others are found.as.pharmacists and engineers. In

contrast over 68 percent of white women in this group are employed

, _,._-_-

in-just three occupations: aeAnimary school teachers, high dchool_

-teachers; and professional nurses.

.

The-white males in the sample average over 14 years of

schooling and nearly 45 percent have_had_some forM of institutional

training beyond their formal education Very few (t1%),,,are members of

- -trade_unions and the a-Verage_minori(ty_employment-im-their occupations_ . -

-ia-oniy 10.8 percent. The white females in this stratum have slightly

less'educption (13.0 years) andunly one-fifth have had any

vocational training outside of formal schooling. Union membership is

weak for women as it is for men (10%) but the average minority

0

202

185

employment in their occupations is three times .the male rate (34 %).

Of total minority employment in,this stratum over 90 percent'are White

women suggesting again the extremely small.proportion of-blacks.in

these occupations. Table 5.7 presents the regression results.a'

White Males - Among professionals and other highly skilled

..

personnel, formal education becomes the primary variable eAolaining. . %

_-.wag.-2 differentials. For white men, each .year of education is.worth1. 1

.

more than twice its value in any of the 'lesser skilled_strata. -An

- 1

additional year of schooling. is worth $.3i an hoUr,and there is AO--1. . _

- ,

_ differential associated with,where the scPolir3_was taken. _Only- in11\

7- -

'this higheit skikled4 stratum is there no discount for_-southern

education. The importance of formal schooling is, of course,,fully4

consistent with the type of-training usuallyrequired for these pto-

__-fessional_occupations.

Migration also plays a role. A $.48_wage differential exists:-

t i

between migianis_ and those whd4lave never moved -from the area in which,

1

they livecl_atp

age 16.' On a_full_ time full-year ,basis, this_is equivalent

,

almost a $1000 annual salary differential. Years of experience 4.p_

alsp especially.important_aading more than $.03 tO hourly earnings per

- -

year of labor force participation. Each year of eXperience translates

r *

-" an annual $60-salary premium. Finally specific vbcdtional\

preparation adds nearly $.70 per SVP unit to the houily wage. Altogether

-the human capital module explains about a quarter of he variance.

The addition of other modules co the equation doe not

appreciably improve the fit. The only significant variable is after-tax

203

o

186

TABLE 5.7

RECRESS/ON EQUATIONS:OCCUPATION STRATUM 15-17 BY RACE AND SEX

a

white Male Black Male

Constant -5.3929 -6.3036

: NUMAN CAPITAL MODULE ,

Schooling .3253 1.3192(7.26) ' (7.21)

Schoo:-Sbuth -Training -Migration -:4123*

(2.01) (1.84).

Experience .0318 .0351- (3.04) (3.37)

S pecific Vec. Prep. .67071 .6934(2.56) (2.68)

STRATIFICATION MODULE_---

===Unfon:Member-

XMinoritp-iIndustry?zt.- --

2-Minarity0ccupatioa1/11a-ckMale-Occupation-

X=Black Feiale--Occupation

-INDUSTRY-MODULE

Concentration

Unicti-x Conc.

--Af ter-tax Profit

CaPit'al/Labor Ratio

-

.0.10

SAMPLE TOO

SMALL FOR

SIGNIFICANT

RESULTS

NOON.,

4.1.04111,

15.5935 --(2.73)

--OW

--eGoverriment Demand -

NOPX2ING CONDITIONSMODULE .Physical DomanNegative Work Traits

2-Ft .247 .267

SEE 1.7809 1.7608

SAN , '$4.83 $4.83

287 287

*Percent Fectale-Oc cispation

White Female Black Female Cross Race-Sex

-.5661

.2620(3.59)

v..1.3813 .

.2853(4.13)

`N.,

SAMPLE TOO

\SMALL FOR

SIGNIFICANT

RESULTS

11.111 .10

-4.7987.-4-A604

.3513 .3196(8.34f (7.79)

-.4913 -.4622(2.20) (2.19.0279 .0320(2.79). (3.30).5322 .4544(2:18) (1.96)

0

--- 1.2463_=-- (222)_

-*7 - =2.47174(3.82)_

=0. 1=4110

. 0

.316 .422

.8954 .8385

$2.e3 $2.83

.7297(213)

14.6214'(2.80)

- 30 30

.204 ;

6

4.111411a .2541.7765 -1-.7110--

$4.63 $4.63_322

/

187.

profits; the tld evaluation is worth $.58 or 12 percent of the mean

wage,for this group. The addition of this factor increases the R2

by merely .02 and reduces the SEE by_only 2_cents. For all intents

and purposes stratification is not particularly responsible for the

variance in 1 earnings among white male professionals. 4' only among ,

1

this special group that the ('relatively) pure hums. , .2ital hypothesis....

holds. ' 6, ,.

-'1 White Females,-7 the.sample of white females is quite small,_but

sohof the vadance in earnings can be explained. In this case,

-

-however, ly education is significant in the huMan capital module

_end this_factor alone is responsible for 32 percent of the variance-in-

earnings". Each yearof formal schooling is worth at least $.26 an= _

k

-hour *hichtranslates into a sizable_9_percent-rate of return based)-

od- .method-ethod_ used throughout this analysis. This high gh rate'of return

s_no aubt,due to the effect of advanced_degrees opening up-access

o_occupations beyond teaching and nursing. Migration, experience,

-ti=

_

ene,0? do not' appear to explain any of the wage differential within

thii high skill strata ofswomen.\

_E- The inclusion bt he indusr ! module adds considerably to'the

explanation of earnings. The only,significant'factor isICOncentration,

--hut this-variable alone raises the coefficient of determination to _.422_

1 \,,

..

, -and-red uces the SEE from $.8954 to $'4-.8385. The ±10 evaluation of the=

_market power factor is valued at $.70 per hour which is equivalent

almost 25'percent of the mean wage for this group. Thus, while the

pure human capital model seems to account for'the overwhelming majority

205

,.188

of explained variance among highSkilled white men, other factors

still play a significant role in explAning the earnings of white

female professionals.

Cross Race-Sex - The cross race-sex equations contain, a small-

---

-number of black men and women,/}as well as whitee. The human capital

module results are similarto those for white men, but=in additionaz/

- 7-

variables in both the stratification and industry modules are significant.

The human capital-/mdaule contains schooling,-migration, experience,

and SVP., Each 'of the e!actors 'has a regression coefficient similar

-thosetin the white male equation. In addi,3; 1, before adding-the

b_ race and sex dummies the percent of female employment in an occupation

---_(%FMOCC) is highly significant and the ±lq test has_a value of4:76

__,anrhour. In the industry module both concentration and aftertaxi-_

___profits_are significant variable0 with the ±la evaluation Worth $.42

-add= _$.54 respectively. The inclusion of these three variables raises-

--the_R_.from .254 to'.315 and reduces the SEE from $1.7765 to $1.71314-

-When race and sex are added, only the dummyvariable for SexA.s,-

significant probably because of the verysmall number of blacks in-

-the subgroup sample. EqUation (V) gives these final reSults.-0

w = 4.9651 + .3120 Schooling - 4031 Migr tion + .0335 Experience

(7.72) (1.90) (3.53)

.(2-.14- .5524SVP - 1.4339 %FMOCC + 1513525

(2.99)

Profit

4) (2.05)

- 1.3944 Sex(3 8) RL 3 .338 SEE = 1.6814

206

,l89

_

The coefficients on the -human capital variables do not appreciably

change'after the addition of the dummy variable to the equation, but

the,coefficientbn %FMOCC falls from -2.47 to' -1.43 and concentration

. becomes insignificant. There'is obviously a large wagdiff ential

associated- with sex per se, yet the "crowding" factor still

significant as does after-tax profits. Given racial and.sexual differ-

-

ences in the labor force, stratification by occdPation and industry

plays some role in wage determination even at the top of the

occupational hierarchy. '

Again, further analysis of the regression results for individuaL

1=occupation-sfrata will be postponed until,the next chapter.

CROSS-OCCUPATION-STRATA _

The evidence-presented to this pant indicates that-Within-

;brbad occupation groups, stratiiiCation and industry variables

t

contribute to an-explanation of existing wage differentials,- In all

-cases these variables are of the proper sign, usually of large magnitude,_

= =- and -have relatively high t-values. Except in the case of the white

-.male equation in 0CC STRATUM-15-17, the addition oLAhe non-qiuman

capital modulles significantly boosts the coefficients Of determinetion

and reduces the standard errors of estimate. We can conclude that

within most occupation stratathe general model of wage'determination--

posited here is superior to any deyeloped in the tradition of pure

human capital or, for that matter, pure institutional theory; _

190

But the more severe test of the relative merits of human-,capital,

institutional, and stratification theory requires evidence across.-

occupational strata- As we have mentioned; it can well be argued that-

the findings-within strata do not ultimately test the theory since

individuals invest in human_capitar )5stensibly 'to move from one

, .

stratum to another. Testing the human capital theory within a single

stratum is therefore biased.in.favor of the institutional and stratifi-

cation hypotheses. This bias is eliminated by pooling the sample across7

-occupatibn-strata. The 'full impact of the human capital module can

% . _

then be measured. Table 5.8 provides these regression results. -

White Males - For the full- time -white male workfoice in the 1967 .

LEO sample, earnings averaged $3.42 an hour,with a standard deviation

2-$1.60---Based on eitheraisiMple R test or based on the changaln-

-the_standard error of the estimate, little additional variance appears-

to=be. explained by variables other than human-capital factors. The

complete equation including stratification, industry, and working=

=condition components increases the by by only .033 and reduces the--

:,.. .

_

---f--=_- --SEE_only slightly.

Each of the human capital factors is statistically significant

for white men with the exception of the vocational training variable:

-Each year of edu4ition is worth $.20 in hburly earnings if the

achooling was taken outside of the south._ Southern- education is valueid-1

at two cents less reflecting only a slight regional differential in

the returns to schooling for the white male workforce as a whole.

According to the rate of return methodology used throughout this study,

208

191

TABLE 5.8

. REGRESSION EQUATIONS:ALL OCCUPATION STRATA BY RACE AND SEX

White Male Black Hale

Constant

HUMAN:CAPITAL MODULESchooling

SChooI-South

Training

_Migration

-Experience

_ Specific VOc. Prep.

STRATIFICATION MODULE'

:Mnion-Member

--s,-

Z-Minothy--Industry

Z Minority-Occupation

Z-Blick:=Maleccupation

_2 white- -Female -- Occupation=

-2-_Black4Femaler-Occupation_:---

_-- INDUSTRY MODULE

:-:Conc-efitration

White Female 131adk. Female Cross- Race-Sex

.1911 .3353 1.5519 1.0838 1.5963 1.6647 - .9224 1.7209 .3113 1.0962 _

0

.2030 .1767 .0911 :.0.721

(15.62) (12.99)' (8.93) (8.05)

-:0199 -.0160 -.0471 -.0315 -.0249 -.0182 .0311 -.0215 -.0260 -.0186

(2.93) (2.37) (7.48) (5.73) (3.61) (2.91) (5:02), (3.77) (4;58) 13;44)

.0578 .0429 .0905 .0749 .1526 .1405.

(3.66) (2.95) (7.13) (7.64) 13.63) (12.66)

--- .2442 .2059--- (2.89) (2.83)

-.2884 -.2861 -.3745 -.2738

'(4.24) (4.28) (6.72) (5.65) ---. _

111

Immaew

.012? .0119 .0065 .0062 -...0095

(3.97) (3.90) (2.41) (2;03) (2.89)

--- .2315 .1620

(3.17) (2.35)

-.2230 -.104 -.3343 --_.30370.010 (3.66) (2.94) .(6.01) -(5.83)

.0059 ;0065.' (2:19) (2.50)

.1596 .1541 .0641 .0883 -.0478 -.0570 .0365*. .0468 -.2076 .1222--_

(8:67)- (8.42)- (3.93) (6.22)__12.00)-_(2.37)!--(1.66)_ (2.3) (13.22)A6:79)-

.;2460 OFR. .5452 *MY .2814 ---- .2850 --=--__:.3760f

, (2.00) (10.46) 1388) "=!.- (4.97)

-.5306 -.6746-_ --- -.5370 ---- -1;0032_ = _

(2.47)_

Uniiin:-x Conc.

-i Profit

f 1

CapitalfLabor Ratio

-_-Cowrnment Demand

-_--WORKING CONDITIONS.

---_-MODIME

...11

- Physical Demands---

Negativt Work Traits

--- 2- R

SEE

MEAN

.23

1.4142

$3.42

1850

(4.23)-- 13.00):-

.11 - 41111.0. a/MOM

.8599

(5.61)-

-.5671(2.49)

6.3688 8.2545 5.9787

(3.26)

4111.

-.1340-

(3.16)

.110-4.

"---;

(5.76)-

.0023(3.90)

1.0711(3.99)

flONIM4. =Mt

011,

12.96)

.0027

(3.02)

2.2590-(2.90)

-.232314.-79)

- -..256 :207. .421 .048 .155

1.3862 .8117 .6953 .9659 .9136

: $3.4/ $2.37 $2.37 $2.05 $2.05

1850 912 912 932 932

209

(6;59) (6-.31)

- ----(6.53)-

O 1.

--.3171 MY(2.64) (5.38)--=--==

=

6.5566

.001.0-

Maw.1

12-.00)1_ =

-(2.22)-_

-.1436 --o;

(3.27)

OaUPO. -4040

.291 ,

.564

$1.66,

397

.431 1 .238,- -.333-_

.5074 1.3270_1.2439 --

$1.66 $2.96-42.96__

397 2394 --2394

'192

4/7

the'average*white male worker reaps a 5-1/2 percent return by

remaining in school for an additional year (at the mean). This clearly

exceeds the rate of return earned by each of the minority groups'in*

the labor force; it is double the rate for black men and more than four

times the rate earned by white women.24

At least in relative terms,7,3

additional education is a good investment for white men, a finding

"consistent wi h virtually all human capital stusties.

Migration, experience, and on-theL-job training_ also play important.

./

parts in the wage determination process. Non-migrants earn, on average,

$.29 less per hour thamthose who have moved at _least fifty miles__1

a

-frOm7their place :residence at-age 16. This_Is_equivalentto_..:-

almost $600 per year for a ,full -time worker. Each year of labor force-_

.k,__

experience adds another 1.23 cents an_ hour to the wage rate. In t

annual terms this implies a $246 differential betwlem the earnings

of a fifteen year labor force veteran and a worker whoAlas been out

:Of school for only five years. -Finally each unit of specific'

_ Notational preparation is worth_$.:16per hour. Given_the full--range

_of this variable, there is a _$1.44 difference In_earnings_betweeda_

-worker in an occupation which requires only a short demonstration

-period and a worker whose occupation requires at least 10 years of on-_

--=:the-job apprenticeship. On an annual basis the impact of ,SVP has a

/24Theactual figures for the four race-sex groups are white =_

men -5.5%, black men 2.75%, white women 1.257,, and black women 4.75%:_

_These relative rates of return are consistent with apriori theory-and-re further explored in the text.

range of $2,880.25

While/the other exogenous

an-explanation of the variance,

are statistically significant.

-- _interact in thenOw familiar ma

1:93

factor modules add only slightly to-

six of the variables in these module&

_Union membership.and concentration

ner.

.

Concentration (MPF).

20%

$3._24

1.37 3.49

+4;-0%_ -2.8%

__These results indicate, thatunion_membership has only a marginal impact-- , _= :

_

-A-'4- --_7z-i- --

___on= relative wages in _both _competitive_and concentrated industries,_ a_

_-_---1

conclusion departing frommany _institutional 'analyses and_roughly

..:-

-:'= consistent with Weiss!su-results. In a_ similar' regression,. Weiss found-- '=- -=:---i-

--___ z , \thatunionization increased earnings -by at most 6=8 Pereentf$5ra-__

-- __

----comparable group of workers.2

---

6

, One example will serve-to indicate the magnitude of thepotential wage differential_based on these regression resulfs. A white

male high school drop-out with ten years 0 schooling_100uover migrated,

has worked in the labor force for five years and is pr'sently employed

--In an occupation which requires only a short tdemonstra on to learn its

basic skills will earn, on average, $2.15 an hour, Alternatively, a

.college graduate who has migrated, has 15 fears of laboA force e:xperi-andence and is presently in -a job requiring between one two years of

on-the-job training will earn $4.58 .pqN hour. This is equal to a $2.43

wage differential, the college graduate earning 113 percent more than

the high school drop-out-.= -,

26Weiss, op. cit., p. 108.

211

I

194

4

Concentration is, more important in the present analysis. Weiss

found a forty poineinCrease in concentration increased earnings by'

_only-3-5 _percent. Hereve-find,the- ncreade to be as large as 10.8

percent in,the non-union-sector. Again,we attribute our finding to

-the better measure of concentration used-in the present analysis. The. vr

weaker effect of concentration on earnings-among organized workers_

implies that unions in.the competitive industries haVe the ability to=y

Win wage contracts more in line with the pattern -set in the oligopoli- _

:stic sector while unorganized Workers in the competitive,sector do

===

not have_ t27

hisopportunity. 'Overall, a unionized-worker In a concen-

c-trated industry earns 7.7 percent_more than a sitilarly=akilled nOn-

union -worker in the classically-competifive sector of the economy._

Two other variables in the stratiic4tion and industry-_ modules

affect white male earnings. A ±la difference in MININD.is _valued

$.16 an hour while a similar ±1 evaluatiOn.,of after-tax profits

implies a $.22 differential. In ;both cases the effect_is statistically

eignifitant, but relatively minor be4qg only4.6'percent and 6.4 _percent_-

of_the man_ wage. In addition to these variables, the physical dem=

.,

factor hat a significant negative sign. Heavier work_apparentli earns --0

.,

I

implication we draw from these results is_thus,,,,at variance- :-

with the ov rall conclitsions of_Weiss. He Orites, "The implicafion=aeets--' _to __be that irms in concentrated industries do pay their _empioyees mo=re,

__-but that th get higher 'quality' labor In the bargain. Theincomes-won--_by union for their members more exteedwhatthose.wOrketa-

. _ ,-- -_

would earn i their_best alternative employments. "__ To the_extent_thait

_it_is possibl to differentiate between the effect ofconcentration:lnd

- -,-unionization, the present study appears to indicate that mOnopolyrepts

_ _arise more fro the produat market structure of an_industry than fro-A

te:present of-Unionization. Weiss,- p. 108..,. 0

0

.

a lowet,wage after controlling for human capital characteristics and,

industry attachment. Tbe coefficient on negative work traits is not

significantly different from zero. In neither case is there an%

indication of a compensatory earnings effect.

-Black Males - The regression results- for black males "are

tosharp contrast to those we have just seen. .'The human capital module

is responsible for less than half of the total explained variance in

earnings with the addition of the stratification and industry factors

reducing the standard error of the estimate considerably. The

essential structure of the earnings generating functions have a

cant __racial= component; as we shall see.

Every human capital factor in the black -regression is signifi=

cant. Schooling taken outside of the south adds $.09 per hour fbr

=every year completed. This is` less than half of the_ increment afforded

-comparable whites and amounts to a rate of return of-less than 3/

-1-

28 _..percent at the mean. Thi. low hourly increment and the in return-

are consistent with virtually all o f the studies that have b een made

of the impact of formal schooling on black maje,earnings.2 9

What is

more, southern schooling is worth only half as raucti as ichobling taken

elsewhere t presumably reflecting the poorer quality of southern black

28The difference in the coefficient on schooling betWeen the

white male and black male equation is significant at considerablybetter than the .01 level according to the difference in means test.e-r-16.77.

4'

29See_Hanoch, op. cit. and Bennett Harrisbft, op. cit. for two

Important studies in this regard.

213

196.

schools: The discount for southern schooling is much greater for

-black than for white men suggesting that the quality difference in

education between southern schools and all others may be primarily

race-related. In the-non-south, the'relative dollar return to

schooling between black and white men is (.0911/.2030)=.45.

south, the equivalent ratio is .24.

Vocational training is also a significant factor in the black

male earnings equation. This is the only group for whiCh this is true

In the

_implying that although institutional- manpower programs_ do not

appreciably affect the earnings of most workers, they do 1;euefit black

meit. Enrollment in a training_ program- is valued at $.24 an hour or

somewhat_ in Sxdess_of 10 percent of-the mean_wage._ Whether these

__programs actually__increase "endogenous" productivity_cannot_be direct.

meaSured,,of course. _What _ig ic'thesnifant coefficient suggestsmaycourse.

=V' _r

_-o4ly be that-black workers uto_have completed a_training program are-

=more likely to be hired for jobs_ that pay somewhat higher wages.= :

Migration is another powerful factor influencipg wages for_this _

_group, For the_black male workforce as_a whole, migration_is_worth

-0:_an_average of $,37jah hour, No doubt:much,of this_overall increment__

_

J

reflects the special beneficial effect of moving out of the south.4

The high rate of return attendent to southern emigration is most

-likely responsible for explaining the higher coefficient on "migration"

-compared to the parameter in the white male equation. Outside of the

south, migration may fail to pay off as handsomely for blacks as

214

197

it does for whites.30

Labor force experience increments earnings by $4065 an hour

per year. Each year in the labor force is consequently worth only

about half the rate for white men implying a much flatter age-earnings

profile. Finally each unit of on-the-job training (SVP) is worth

$.06 an hour. This figure too is less than half the coefficient for

white men. Part of this difference may be the result-of unspecified

non-linearities in the return to specific vocational preparation.

Alternatively, the smaller coefficient indiCates a real difference

the return to leach Unit_of SV_.

Taken together the_six human capital factors explain one-fift

_

-of the variance -in earnings among full-time black male workers. Thy- -, --:,=-_ -_

_addition_of the three remaining modules increases the coefficient-_ f

--_determination to .421. Union membership is extremely powerful_in_the

complete equation. The nearly $.55 wage differential_between union

_and_non-union work- ers represents an average_union wage which 25.7

_

percent greater than that received -by the average non-union worker.

ipusly exclusion from a_trade union has_a massive impact on the_ _

Evidence for this statement can bel-oundin Barry_ Bluestone,_William Murphy, and MarySIVenson,-Low- Wages and the Working Poor- _4

Unn.Arbor: Institute of- Labor and Industrial Relatiniversity,_ofMichigan-Wayne State University, 1973)=, p. 127. Regarding black males,_

-

-- "Mobility out of other-regions_of the nation (other than the_ -_ south) does not pay as handsoMely. Across-all-education_

-groups, moving out of the-Northe'abt is only slightly bene-ficial _fox those who move. to the North Central states or to

the West. All other moves actually increase the probabilityof pOor paying jobs."

215

earnings of a black male worker. While not a particularly important

factor for white men, unionization represents a most important route

to higher pay for the black male workforce. This is consistent with

both institutional and stratification hypotheses. The percent of

minority employment in an industry (%MININD) also affects,the_-

earnings distribution for this group of workers. A ±la difference

in XMININD is valued at $.22 an hour, Just slightly higher than the

effect on white male earnings.. I

The industry module in the filial equation has a structure-which___--

basically different from that-of white men. Neither concentration- - -

nor the interaction term are reported in the final equation, although

n_test runs concdntration (but not the interaction term) was

extremely significant and powerful. It was necessary to_drop concen7__-

tration from the final equation because its addition always destroyed

e_integrity of one of the human capital factors. All_of the

regressions which were prepared with concentration as one of the

_exogenous variables failed to include n experience It as a statistically

significant himan capital- factor. It was impossible to pin down the

reason for this deteriorating effect on the "expe'ience" coefficient.

As a substitute for concentration, other dustry variables

were significant in the complete equation with ut harming the_human_

capital module coefficients. These included the highly colinear after/

tax profit rate. The ±1Q evaluation of this variable is worth $.32

= an hour. Simp.ar ±1Q evaluations of the/capital/labor ratio and the

government expenditure variables are worth$.18 and $.20

216

199

respectively.31

Each of these effects takenIindependently have more

than a minor impact on the distribution of earnings. To the extent

that these effects are additive,-the industry module is quite powerful.

The case for "complex" crowding is convincing while the human capital

explanation leaves much;to be desired.

I

The actual importance of human capital in explaining'the -, \

existing wage differential between white and black men can be quanti-.

'fied by &sing the information generated in the regressions. The1

average wage for black male workers in 19_67 was ,$2.37 or 69 percent

of the average white male rate. The standard deviation was $.91.32-

\_

These three industry factors make perfect"quasi7"instrdmenta_

variables. They are colinear with concentration-but not_=with_ _

in-the human capital module. The partial (XtX) matrix for the trelevan,_

-- -factors- is reproduced-below. _ _

Partial (XtX)_Matrix=for Black_MaleaCross-Oecupation Equation

Concentration Schooling -Uperience-

after -tax Profit .5470 .-1464 -.0797:_ -

=Ratio .3706 --0614 .0758_ --

vement Demand .

1

Go rn 1352-1== *.0383 ..0069--_,

I

_ _

Variability-in-earnings in the sample white male population- -ig'nsiderably greater than in the black male group. The coefficient-of=Variatiol (V) for whites is .4678 whileIonly .3838 for the black,

'-f,

_male sample. Two factorsmight eXplailt.-this difference. One is that _

-the Underlying black male population _1. more homogeneous in human_-_capital and therefore more homogeneous in earnings. The other=-is- -that

the labor market treats black men as though they were more homogeneousin human capital than they really are (i.e. employers disregard human

--capital differences Or discount them). In the first case we wouldexpect to find a greater V forjlle-Ahuman capital characteristics of

,,: r_ a

A.or),A. 7

-200

If we subL.Zitute the black means into.the 'te equation the hourly

rate for black males rises to $2.58 or 75 percent of the white male

mean. Furthermore if we substitute black male means for the human

capital module, but, white male means'for.the other modules, the black

male wage rate increases to $2.73 ors80 percent of the WM average.

Assuming that SVP is a stratification variable'becduse it_is acquired

_ on the job after access to employment has been secured, the black male,

Wage now rises to $2.97 an hour or 87 percent of the white male mean.'

In this certainly plausible case factors other than human capital

account for over 56 percent of the BM/WM.differential and only_43-1

;

percent of the mean _wage difference between white and black males is

the white male-group.we find- the opposite to be true. Fot-each-of-the

jaitan_capital variables with theexception_of School-south,_thalr's-__Eor-white and black men are generally equal or the coefficient is

greater fot black men.ra

V .

bm-

Earnings

SchoolingSchool-SouthTraining-MigrationExperienceSVP

.4678 .3838 32.2188

.2773 - .4011 .6913

1.7467 ..8480 2.16171.6995 2.5823 .6581

1:2046 1.1601) .1.0383

14344 *.4220 1.0293

.3610 .4582 .7878

_This implies that the labor market is_leas sensitive to differences inendogenous productivity characteristics of the black male workforce.

-Larger relative variability in education, for instance, is notreflected in the variability in earnings. This does not necessarilyimply that individual employers who hire blacks totally overlook

ences in worker characteristics when choosing their_ employees. But-it.does provide another cogent piece of evidence-that the_labor marketlacing black workers is substantially restricted.

218

201

Atdue to measured differences in causally prior human capital yariables.

33

0

These results are summarized in Table 5.9.

TABLE 5.9

POTENTIAL WAGE RATES FOR BLACK MALE WORKERS UNDERVARYING ASSUMPTIONS ABOUT THEIR CHARACTERISTICS

Assum tions

(WMr-=$3.42)wm

Wa e BM/WM Ratio

MHC,Strat,-Ind, WC

--11-11111C (-SVP)

wmStrat, Ind, WC, SVP

$2.37

2.58

.69

2.73- .80

2.97 .87

BM = Black male estimating equation_WM = White male estimating equationbm = mean values for BM exogenous variableswm = mean values for WM exogenous variables

33The differential due,to measured human capital factors i&

calculated from:

b-MHC(-SVP)WM - WM--

HC

wm --Strat, Ind, WC, SVP

WM= - BM

219a

202

White Females - The overall structure of the complete white

.female equation is somewhat similar to that for the black male

workforce with two important exceptions. The first is that neither

the human capitalregresiion nor the complete equation are very gdbd

models of wage determination based on the coefficient of determination

or the standard error of the estimate. The second exception is that

the human capital,equation contains neither-training nor migration,,,both

of-which were significant variables inthe black male regression.

addition the sign on the experience variable is negative.

The unaugmented human capital equation explains only 4.8 percent

Of the variance in white female earnings and each'of the exogenous_

-variables is relatively weaL A year of_schooling in the non -south

=worth less than $.06 an,hour at the mean while a year of southern= =

schooling:is worth.only $.03. In the nonsouth this Is equivalent

to----a minuscule 1.25 percent rate of return on a year of educatioh, the

'lowest for any race-sex group. Based on this evidence, schooling ddea

not generally appear to be a very profitable investment for white

--women in terms of their own future earning:;. Vocational training is

not very profitable either. Although almost 11 percent of the sample

_had some form of institutional training, enrollment in such programs

does not have a significant-impact on earnings. As we have mentioned

previously; according to our regressions, oflly black men earn more due

to manpower programs. Migration plays no role either. This was not'

unexpected given the assumption that men migrate for economic reasons

while working women generally follow their husbands rather than seek

220

203

to maximize own earnings through geographical relocation.

When running the human capital variables alone, a negative sign

is found on the experience variable implying that more experienced

women earn less given equal years of schooling. As we noted earlier,

this result may be illusory because of measurement error. Given the

pattern-of fdmale labor force participation the "experience" variable

does not accurately measure the number of years in the labor force.

_However if human capital "depreciates" with mon-participation, it can

be expected that a woman who returns to the labor market after a period

of time out of the labor force will earn less than a woman who never

left work. This could explain a flat or negative earnings profile

with respect to the variable "experience" or, to be more accurate,

age. In the complete equation, the coefficient on "experience" is- not

significantly_ different from zero indicating a flat "experience-

earnings profile after controlling for all other measured factors.

Specific vocational preparation is barely significant at the--

.05 level. Each unit of SVP adds less than five cents to earnings,

an amount smaller than a third of that in the white male equation.

Again the relative size of the female coefficient may be biased

downward because of non-line;-:ity in the variable. But this seems

unlikely to explain such a large difference.34

34,Alternatively, the weaker earnings effect of on-the-job

training found in the white female equation may reflect a significantinteraction between this variable and other human capital factors.It can be hypothesized that each additional unit of SVP in combinationwith education or other human capital factors has a higher rate of

221

204

The addition of the three remaining modules increases the K2 to

.155 and reduces the SEE by $.06. As in the black male equation,

inclusion of concentration only came at the expense of violating the

proviso concerning the human capital module. Coefficients on both

schooling and SVP fell significantly when concentration was added to1

the equation. Consequently other industry variables were used as

quati-instruments.

Both union membership and MININD were significant in this

equation. Union membership is valued at $.28 an hour leaving organized

-workers earning 14.2 percent more than non-union employees. The dollar

,amount is approximately equal to that of white-male workers but wily.

-_about-half that of black men. in addition the ±la evaluation of

has'a valUe of $.22 an hour around a mean of $2.05. As in the black

male equation, after-tax profits, the ,capital /labor ratio, and Wet:,

government expenditures variable are all significant.

±la Evaluations

After-tax profit rate $.20

Capital/labor'ratio .19

Government demand .18

"Additive" Total $.57

return. Without some form of complementary investment, SVP alone is

worth little.Given the lower mean SVP for white women (SVP =5.28 vs.

SVPwf=4:16) this could explain the difference in the coefficients.

To test this we ran an interaction term including7GED and SVP and

another with schooling and SVP. Both variables were insignificant.

The lower coefficient in,the white female eqqation apparently either

represents the effect of specification error or implies a significantly

lower return to on-the-job training. The coefficients on SVP for the

white male and white iemale equations are significantly different at

better than the .01 level. (t' =3.96)

222

205

Finally the physical demands factor is significant, but once again

negative. Physically demanding'work is rewarded with lower wages,

other things equal.

Compared with the other race-sex groups, relatively less of the

variance in white female earnings is,explained' by the general model of

-wage determination. The use of interaction terms in the human capital

module might have improved the fit, but experimentation with these-

variables proved fruitless. Apparently there are numerous other0

factors not taken into account in the model which have special relevance

for whits women.3

35Conjecture-leads us ,to believe that one set of factors

determining earnings not taken into account in the general model relates

to the importance of earnings for women in various types of house-

holds_t__-Ceteris paribus, a Woman's earnings may-be inversely

relged to her family's ability to provide a sufficient income to

keep the family at a "satisfactory" or target standard of living. Where

the =woman's earnings are an important portion of the family's total

income, wt might expect more intensive job search by the female in

the household with earnings being the key argudent in her utility

function. Earnings may be a much less crucial factor in job choice'

in= families with sliffirient income from other household members or

alternative sources. In this case, two viomenvith equal endogenous

prOductivities may earn significantly different wages.Another set of factors that may be important in the earnings-

function for white women has to do with physical appearance and the

7 production of "psychological" benefits to employers. according to

/ Paddy Quick, women may be hired for other reasons than objectively

/ measured productivity; they supply their bosses (and their customers)

with a more or less pleasant social and psychological environment.-

The human capital characteristics measured in the present study may

not capture the-traiti which are "productie" in this respect. With

these factors missing, the general model fails,to account for a large

part of the variance in white female earnings. See Paddy Quick,

"Women's Work," Review of Radical Political Economics, Vol. 4, No. 3,

July 1972:

223

206

Although our equations leave a good deal of thewariance in..

earnings unexplained, we can, still estimate the impact ofILV)le human

capital module on the wage differential between white women and men.

This can be done as in the black male equation by varying assumption's

about-the mean values of the white female exogenous variables. The

results indicate that human capital is an extremely inadequate explana-

tion of the forty percent wage gap between white men and women.

.Plugging all of the white fetale means into the earnings

equation for white men increases the WF/WM ratio from .60 to .90. ,If

we use the white male means in the stratification, industry, and

---=t=Working conditions modules and the white fetalehuman capital4wans4---

thel ratio_ rises to .93. -Finally if we assume that- SVP is ,

-cation- factor rather than a lumen _capital variable and we, evaluate

- the_-white male -equation once_ again, we eliminate practically all-of

the difference in earnings between The two_ groups. Only .42/.40=5

__percent of the differential is directly due to sex=related-vdifferences

-In schooling, training, migration, and "experience." 'Given the

-Measurement of experienCe this may be a slight underestimate of the

*

full impact of human capital; but the thrust of the result still

stands even if we discount this variable by a large percentage. The

huge wage difference between white men and women cannot be attributed

to the latter's underinvestment in human capital. Crowding and other

forms of labor market discrimination play a much more critical role,

although other factors not inclUded in the model may be most

important.

224

207

TABLE 5:10 -

Aot

POTENTIAL WAGE RAGES FOR WHITE FEMALE WORKERSUNDER VARYING ASSUMPTIONS ABOUT THEIR .

CHARACTERISTICS

(WM = $3.42)wm.

Assumptions Wage BM/WM Rati

Strat, Ind, WC $2-.05 .60

WM--wf

HC Strat, Ind, WC 3.09 .90

-wfHC

Stra t, WC 3.18 .

==-HC(-SVP)

w`°- Strat, Ind, WC-, SVP 3.36====

Black Females - Black women are by far the poorest paid members

of the workfOrce. With an average wage Of $1.66 an hour in 1967,

black women earned only 48.5 percent of the average-wage for white

men and 81,percent of that for white women. Unlike white women, however,

the general model of wage determination_ is capable of explaining a

-good portion of the variance in their earnings. The human capital

.module alone is responsible for 29 percent of the variance and the

Complete equation has a corrected R2 of .431, _the'highest Among the

four race-sex groups.

Schooling plays a much more important role for black women

than it does for either of the other minority groups. This is primarily

225

208

due to the impact of edudation on occupationalmobility.36,

A year

of additional schooling (at the,mean) in 'the non7south'yields a wage_

increment of $.09 .°hour; in .the 'sduth, .06. This is more than0

\\

fifty percent higher than foi white women and, equal to the wage

increment for black men, Because of the extremely low opportunity costI

of additional schooling, the rate of return for black women is only_

secbnd to-that of. white men. A marginal year of schooling yields a._

4-3/4 percentrrate of return, 'only 3/4 of a percentage point behind

_ .

the white male rate.

Neither institutional training nor experience are significant in

thia--equation. But the coefficient on Migration iuggests-southern

--emigration is useful for black women-whether-main Motive for0

elocation is directly economic or not. A black woman who relocates

___ _ earns, on average,,P14,-2percent more,($.22) than _a similar Worker__

-- /_

co never moved more than 50 miles from her childhood home.

fl0

Spetific vocational preparation is not_ significant (even at the

-.05_ level) in-the human capital.equation. After Controlling for

industry Characteristiei and union membership, however, SVP becomes

0

significant at the :01 level with each unit of on-the-job training

yielding approximately the same return as for white women ($.047). .

The addition of the three remaining modules increases the R2 to

0

.43r and reduces the SEE to $.507. In dollar terms, union membership

36Fora more detailed Analysis of this poiut, see Bluestone,

Murphy, and Stevenson, op. cit.

p

2 26

A,

209

is worth $.29 an hour, an amount equivalent to that for both white

men and women. Because of lower average earnings, memberehip is

valued at 18,percent, more than four times the value for white men and

27 pei,eentmore than for white women._ Segmentation into minority-_

impacted industries isalso inch more iiportant for black women than

for any of the other groups. The±lo.evaluation of %MININD is-valued

--_at_$-.38 an hour, almost twice the_ effect found elsewhere. This is -_

fully consistent with other data which suggest_that black women have

historically been segregated into-a very small number of industries

and-occupations; many of which are related'to domestic and-personal

, _service. The one_ significant industry variable-in the final equation

s--Ooncentration; here the ±la evaluation is worth-$.16 an-hour.,

Together, union membership, -%MININD, and con_ ntration are-Worth_$:-.-83

an_shour, exactly half of the mean wage rate.

Evaluating the white male equation at the black female means

,_-urnishes added evidence of the qualitative difference in the earnings:

functions between the two groups. When the white male equation is

evaluated with all of the black female means, the wage, ratio rises

steeply from .485 tc .75. In this case the higher wages for black

women would be due to the higher gross returns on their human capital

and_the smaller impact of being assigned to minority dominated

industries. (See Table 5.11),

If black women were to gain access to the-same set of industries

as white men the wage ratio would rise still further to .84. In this

Case, human capital differences would be responsible for .16(1-.485)=31

210

percent of the total, wage differential. The other 69 percent would

be due to differences in the structure of the earnings functions

-(varying gross returns) and differential access to industries and

occupations. If we then assume that SVP is a stratification variable-,ii

//

the difference in human capital endowments is left to explain only1I6ri

, _. ,/

percent of the total wage differential. This isa-gpod deal mor

than for white women but substantially less than for black men.

TABLE 5.11

POTENTIAL WAGE-RATES FOR BLACK FEMALE WORKERS UNDER.VARYING ASSUMPTIONS ABOUT THEIR-CHARACTERISTICS /I

(WM==$3.42)Assumptions 'Wage

HC, Strat,'Ind, WC $1.66'

HC, Strat, Ind, WC 2.55

TitRC

wmStrat, Ind, WC 2.86

it

.75

.84

WMStrat, Ind, WC, SVP 3.14 .92

_All of the minority group results thus point, overwhelminLL;

to the importance of factors other than human capital in explaining

the large wage differentials between groups. Differences in schooling,

institutional training, migration, and experience can explain only

two- fifths of the differential between white and black males, only a

sixth of the BF/WM differential and only a twentieth of the differential

between white men and women. The remaining portion of the differential

ePriOfo4

211

is due to a combination of stratification mechanisms: unionism,

"crowding," and pure wage discrimination.

The relative unimportance= f h an capital differences may be due

in part to the specification of the pooled regresSions. The absence of

a log linear dependent variable, and interaction terms for education,

experience, and training may be responsible for this result. But other

investigations come to very similar conclusions as ours using different

.

techniques and.data sources. Blinder's study of wage discrimination

-uSingMichigen Survey- Research Center data cOncludesithat the amount of

Tintergroup wage differentials which can be explained by differences in

personal endowments is even smaller than that fdlund in the present _-

-37-study.- . For _the male_vage differential; Blinder_ concludes that Only

30-percent can be attributed to differences in endowments while virtu-

:01.ynone of the white male/female differential is due-to these

faCtors.38

Using -still different techniques-, both - Michelson and

-Siegel have also-questioned the importance of human capital endowments_;

in explaining white/black income differences.

:THE--1GRAND" POOLED REGRESSIONS-

--. The final three regressions reported.in this chapter are for

39

the total fulltime fullyear privately employed labor force. Even" if

37Alan S. Blinder, "Wage Discrimination: Reduced Form and

Structural Estimates" Journal of Human Resources, Fall 1973.

38Ibid., pp. 447, 449.

39See Stephan Michelson, Incomes of Racial Minorities (Washington:

The Brookings Institution, 1968) unpublished manuscript; and Paul Siegel,

"On the Cost of Being a Negro," Sociological Inquiry, Winter 1965.

74r7046 *a

212

somewhat imprecise due to their specifications, these equations do

clarify the dimensions of "crowding." The numerous caveats regarding

their interpretation have already been discussed.

When regressed alone the human capital module explains 24 percent

of the variance in all earnings. A year of schooling is worth $.15 an

hour ($.127 in the south) which translates into an average 4.5 percent

rate of return on the foregone income opportunity cost of schooling.

The training variable is.significant with enrollment in an institutional

vocational program worth over $.23 an hour.' Migration is worth $.33

;while each year of experience is*valued at nearly 6/10 of a cent and

:_-- each -unit of SVP adds $.21 an hour. The-mean wage for this 1967

composite sample is $2.96 with a standard deviation of $1.52.

Using this equation it is possible to estimate the range in

earnings under different assumptions about schooling; SVP, training, and

experience. For simplicity we assume throughout that schooling was-

taken outside the south("school-south"=0) and that migration had teen

undertaken ("migration"=0). These results are reported in Table 5:12a-c--

along with the estimated earnings for each of the individual race-sex

-Agroups calculated from the own occupation-pooled regressions. The

row W* in this table refers:to the estimated wage for the "grand" pooled

regression.- All of the estimates- are made from the human capital

_ equations reported in Table 5.8. The four rows below the dollar esti-

mates give the percentage differentials. from the grand pooled wage for

each of the race -seat groups. In all but a very few cases, white men

have-wages in excess of the grand pooled estimates while each-of the

o

230

213

TABLE 5.12a

ESTIMATED HOURLY' EARNINGS UNDER VARIOUS ASSUMPTIONSCONCERNING THE HUMAN CAPITAL. MODULE IN THE

OCCUPATION - POOLED, REGRESSIONS:

SCHOOL COMPLETED=8 YEARS

School

Completed 8 Years

SVP 2'

Training No Yes No Yes

Experience,5 20 5 20 5 20 5 20_-

$1.98 $2.07- $2.21 _$2.30 $2.81 $2:90 $3.04 $3,13 :

2.20 2.38 2.20 2.38 :2.84 3.-02 2.84 3.-02=_

2-.41 2.55= -2.65 -2.57 2.67- 2-.80' 2.-91

2.15 2.15 2.15 2.15 2.34 2.34 2.34 2.-34 __-

1.72 1.72 1.72 1.72 1.87 1.87 1-.87- 1.437

% '% x x % % %

1.1*-1.1._-)/W*-411.1 +15.0- --.5 +3.5 +1.1 +4.1 -6.6 -3.5_

ic7Vibm)/ii* 416.7 416.4_ +15.4 +11.3 -8.5 -7.9 -7.9 -7.0

W*-W')/W* 48.6 +1.9 -2.7 -6.5 -16.7 -19.3 -23.0.-25.2

'bfVW*-13.1 -16.9 -22.2 -25.2 -33.5 -35.5 -38.5 -40.3

231

214

TABLE 5.12b

ESTIMATED HOURLY EARNINGS UNDER VARIOUS ASSUMPTIONSCONCERNING THE HUMAN CAPITAL MODULE IN THE

OCCUPATION- POOLED REdESSIONS:SCI400L COMPLETED=12 YEARS

School-Completed 12 Years

SVP_ 2 6

Itaining No Yes No Yes

Experience 5

W* $2.59

3.01wm

bm

wf

2.67

2.38

2.08

(W*-Wwm

) /W* +16.2

(W*-Wbut)/W* +3.1

(W*41wf) /W*-8.1

(W*-Wbf)/1.1*

20 5 20 5 20 5

$2.68 $2.82 $2.91 $3.42 $3.51 $3.65

3.19 3.01 3.19 3.65 3.83 3.65

2.7/ '2.91 3.01 ' 2.93 3.03 3.16

2.38 2.38 2.38 2.57 2.57 2.57

2.08 2.08 2.08 2.23 2.23 2.23

% % X % %

+19.0 +6.7 +9.6 +6.7 +9.1 0.0-

+3.4. +3.2 +3.4 -14.3- -13,7 -13.4

1 -15.6 -18.2 -24.9 -26.8 -29.6

.-28.5 -34.8 -36.5 -38.9

20

$3.74

3.83

3.27

2.57

2.23

+2.4

-12.6

-31.3

'-40.4

215

TABLE 5.12c

A A

ESTIMAtED HOURLY EPRNINGS UNDER VARIOUS ASSUMPTIONSCONCERNING THE HUMAN CAPITAL MODULE IN THE

OCCUPATICN-POOLED REGRESSIONS:SCHOOL.COMPLETED=16 YEARS

SchoolCOmpleted

SVP 2

16 Years

6

Training No Yes No Yes ,

Experience 5 20 5 20 5 20 5 20

W* $3.20 $3.29 $3.43 $3.52 $4.03 $4.12 $4.26 $4.35

W 3.82 4.00 3.82 4.00, 4.46 4.64 ,4.46 4.64wm

Wbm

3.03 3.13 3.27 3.37 3.29 3.3,9 3.52 3.63

Wwf

2.61 2.61 2.61 2.61 2.80 2.80 2.80 2.80

Wbx 2_44 2.44 2.44 2.44 2.59 2.59 2.59 2.59

% % % %

(W*;Wwm)/W* +1.9.4 +21.6 +11.4 +13.6 +10.7 +12.6 +4.7 +6.7

(W*-W, )/W* -5.3 -4.9 -4.7 -4.3 -18.4 -17.7 -17.4 -16.6.,m

-18.4 -20.7 -23.9 -25.9 -30.5 -3,-0 -34.3 -35.6

(W*-Wbf)/W* -23:8- -25.8 -28.9 - -37.1 -39.2 -40.5

233

.

216

minority groups falls below the respective grand means. Black males

with 12 years or less of schooling and little on-the-job training

comprise the one major exception to this rule. ,There is also a.general

trend for the wages of minority groups to fall further behind W* as-the

amount of SVP, training, and experience increases. This trend is less

pronounced for increases in schooling. This all reflects the lower

' earnings elasticities (w.r.t. human capital) prevailing for minority

groups in the economy.40

40It is tempting to .interpret .1* in Table 5.12a-c as the

wage rates that would prevail for gives numan capital endowments in

the absence of "crowding." But this interpretation is not correct

except extremely restrictive assumptions. For W* to be theperfktly competitive ("uncrowded") wage, (1) the underlying distri-bution of, human capital must be identical for each of the subgroupsand (2) the ratio of the slopes of the sectoral demand curves mustbe inversely proportional to the employment ratio in the previously

segregated sectors. _ Proposition (1) is required in order for thegrand pooled regression estimates of W* to equal the weighted meanwage' estimates summed over the race-sex subgroups 6rs) . Proposition

/

(2) follows from the theory presented in Chapter III. The proof of

this is straight-forward.0 I-

Let (1) wi = al-biEl

(2) w2

= a2-b

2E2

and (3) w**= Wrs (w1yw2E2)/(E1 +E2)

with the first equality in (3) holding only if the human,-apital distributions are identical.

If wl = w2 in perfect competition, then from (1) and (2),

(4) El = (a17a2)/bi + (b2/b1)E2.

If a1= a

2equation (4) simplifies to the familiar inverse ratio'

(E1 /E2) = (b2/bi)

Substituting El = (b2/b1)E2 into (3) then yields

217

The addition of the three remaining modules boosts the toto

;333 vial a large number of significant variables. In the stratifica-

tion module both'%MININD and %MINOCC boast highly significant negative

coefficients. The usual ±la evaluations yield wage differentials of

$.38 and $.50 respectively. Union membership interacts with concentra-

tion to render the following effect:

No union

Union

Concentration

20% 60%.

I

$2.79,

$3.08

3.04 3.07

+9.0% 0.0%

In competitive industries, union members earn approximately 9 percent

more than workers who do7not-belong to a trade union, But in the

oligopolistic sector, union membership has no particular impact on '

'rlative. earnings. It seems reasonable to believe that the small ,

ffect reflects the relative extent of trade unionism in different

arts of the occupational hierarchy. With every few highly paid pro-

essionals and technicians in occupations with organized trade unions

as usually defined, the cross occupation union variable tends to

underestimate the impact of trade union membership in specific occupation

W* = f(b2/bi)wl+w21/[(b2/b1)+1]

This reduces to W* = w, = w2

in perfect competition.

Without explicit knowledge of labor demand in each sector it isimpossible to determine the wage impact of desegregation.

235.

218

and industry groups. Indeed, if who belong to professiodl1

organizations which behave like trade unons (e.g. the American

Medical Association) were assigned a duit9y value for membership it

It

seems likely that the impact of the union variable would be

greater in theseequations.

1

Concentration itself is relatively powerful in the non-unionA

sector, but again in unionized indusrries greater co centration does

not translate into additional higher earnings, yet, overall, a union

member in an olignoolistic ind7rry earns 10 percent more than an

unorganized worker in the'competitive sector/of the economy.

All of the other industry module variables are significant as

well. The ±la evaluations of after-taXprofits, the capital/labor

ratio, and government demand are worth $.23, $.13, and $.12 respectively.

Together they play a not insignificant role in explaining eAisting wage

differential7. even after controlling for the effect of concentration

and union membership. Finally, the physical demands factor is sigaifi-

cant but once more negative.

Adding the dummy variables for race and sex to this equation-pro-

duces some further insights. While most of the coefficients on the

human capital variables remain unaltered, the statistical integrity of

"training" is compromised no matter when the dummy variable for sex is

added. The simple correlation between sex and training is relatively

small (-.145), but apparently multi-collinearity between several

variables in the human capital module and sex is sufficient to produce

this result.41

No matter what the specific reason, however, there is

41 Investigation of step-wise regression results on the grand

219

enough other evidence to conclude that part of the explanation for

lower earnings among womer, is the result of less vocational training.

a

w = .8519 +0.1360 School - .0146 SchOol-South + .1010 Training*

(12.71) (2.81) (1.53)

- .2750 Migration + .0078 Experience + .1296 SW.

(5.50) , (3.12) (7.41)

- .3438 %MININD + .0423 %MINOCC* + .3161 Union Member

(2.16) (.28) (3.42)

+ .7738 Concentration - .6090 Union x Concentration

(6.03) (3.47)

+ 6.7310 After-tax profit rate + .0009 Capital' /Labor ratio

(4.09) (2.25)

+ 1.0652 Government Demand - .1389 Physical Demands

(2.71) z (3.72)

- .2920 Race --1.0648 Sex 1

(3.20) (14.29),2

R = .387 SEE 1t1922

Over a quarter of the white male workrs in the sample had some form

of institutional training during thei 'work careers. In contrast only

10.percent .of the white women in the sampseand 14 percent of

the black women reported institutional training.

Of even greater apparent interest, 'addition of the dummy - variables

severely reduces the, coefficient on the industry crowding variable

and totally eliminates the significance of the proxy for occupatio'al

=

segregetien. The coefficient on %MININD falls from -1.0032 to -.3438

pooled eqUation indicates that the multicollinearity apparently arises

between training, sex, and SVP. In an equation with just school and

sex, training has an F-value of 9.07 if entered as the next variable

in the regression. If training is to be added to an equation with

school, sex, and SVP, the F-value for training (if entered) falls to

2.80, well below the F required for statistical significance.

220

while its t-value drops from over 6.3 to less than 2.2. Meanwhile

the coefficient on %MINOCC declines from -.9075 with a t-statistic in

excess of 6.5 to +.0423 with a -paltry t of .28. At first glance

this suggests the near total absence of "crowding" after controlling

for "pure }discrimination."

Combined with other information, however, this conclusion seems

to be much more tenuous. Evidence from the (XtX) . mArix for the "grand"

pooled regression combined with the highly significant coefficients one

XCININD in virtually every one of the individual race-sex equations

strongly ,hint that (1) industry and occupational crowding is widespread

and that (2) workers ihminority-crowded industries are paid less

regardless of race-and sex.

TABLE 5.13

PARTIAL (XtX) MATRIX FOR "GRAND"POOLED REGRESSION

Race Sex %MININD /MINOCC

Race

Sex

%MININD

%MIN=

1.000 .006

1.000

.034

.470

1.000

.051

.637

.389

1.000

The complete elimination of %MINOCC from the cinal equation is

most likely'the result of collinearity with the better measured variable

for sex. In effect occupational "crowding" appears to be so complete

that it is impossible to independently measure its earnings effect.

While there is then no definitive prbof for

2a3

ontention that

221

"crowding" bears. much of the responsibility for the large wage

'differentials ound after controlling for human capital, the mass

of evidence points strongly in thisdirection. This concluion is. / .

reinforced by our,previous findings of a significant coefficient on

%MININD in a large majority of the individual equations, particularly

in'the lower occupation strata. Table 5,14 summarizes all :of the

%MININD results. If the bulk of these had been insignificant, we would

have been much more hesitant to conclude that crowdidg plays a critical

role in wage determination.

f

TABLE 5.14

SUMMARY OF SIGNIFICANT %MININD FACTORS(t- values in parentheses)

Race-Sex Group 1-3 5 6-9 12-14 15-17 Total

WM -2.0851 - .8005 - .7850 - .5306

(4.85) (2.91) (3..18) (2.47)

BM - .7213 - .5723 - .8673 na - .6746

(2:86) (2.34) (2.14) (4.23),

WF - .8593 -.5370'

(2.97) (3.00)

BF -1.4104 - .5404 na = lia' - 1.0994.

(4.21) '(1.96) (6.59)

Cross ..1.1829 -1.1728 -1.2192 -1.0525 -1.0032

(4.03) (6.22) (4.37) .(5.25) (6.31)

Cross w/R,S -1.0019 - .4138 - .7200 - .8550' - .3438

(3.51) (2.13) (2.90) (4.29)' (2.16)

Ort_I979

0

222

,Beyond this, the addition of the race and sex dummies to the

"grand pooled" regression strengthens the impact of the demand side

variables. Coefficients on concentration, after-tax profits, and

government demand increase after the dummies are added and the negative

coefficient'on the union-concentration interaction term declines.

Finally we note that addition of the race and sex variables

raises the to to .337, over 6d percent more than the coefficient of

determination for the human capital equation alone. Clearly then,

human capital is an important element in Wage determination for the

.whole labor force, but the story is much more complicated than all

that. This we shall see-even more clearly in the next chapter.

240

CHAPTER VI

AN EVALUATION OF THE REGRESSION RESULTS

In the previous chapter we reported all of the regression results'

and presented a preliminary analysis of each of the significant

variables. This analysis demonstrated the significance of institutional

and stratification factors in the determination of earnings and

provided substantial although not incontrovertible'evidence.of the0 -V

earnings impact,of inddstry and occupational segregation. The present

chapter extends this,analysis by estimating the overall magnitude of

earnings differentials for (a) workers who share the same human capital

but differ in industry and occupational attributes and (b) workers

who differ in human capital but.work in similar industries and

occupations. Instead of using an ad seriatum analysis of variables

as in the former.chaPter, the present evaluation considers the

variables in each module as a unit (or ad conjunctum). In this way

the combined impact of labor supply-restrictions can be measured as

well as the combined effect of the demand-side of the market. The

results confirm the significanceAf non-human capital factors for

virtually all members of the labor force and-especially for minorities

and all those on the lowers rungs of- the skill hierarchy. As in the

previous chapter, each race-sex group is separately analyzed concluding

with an investigation of the total labor force:

241

223

,The Methodology

There are a number of methods that could have seen used to

estimate the relative strength of human capital and no -human capital

' factors as determinants of personal earnings. A brief r view of some

of these and the reasons for discarding the traditional one serves

to introduce the multivariate method finally chosen for this eurpose.

The simplest method is probably an R2comparison or F-tes

Giv.en the nature of the regression procedure it is easy to measur

how much additional variance in earnings can be explained by the

inclusion of the stratification and industry variables. But it is

not' really the explained variance we are after. Instead we are

seeking an indication,of the size of- potential wage differentials

associated with the non-human capital factors. An R2

comparison or

F-test says nothing about this and therefore is-inappropriate.

The traditional elasticity measure used in most economic analysis

is somewhat more appropriate, but it too has a number of problems

' I

which cause us to reject it in this case. For one thing, point

elasticities may tell very little S)bout the relationship between a

particular pair of factors when evaluated at:points other than. the

mean. Constant elasticity measures can surely be calculated, but they

may bear little resemblance to the real, relationship between variables

when evaluated near the tails of the distribution. While this is a

relatively weak argument against the use of elasticity measures, -

there are additional arguments which are more cogent.

242

225

. It makes sense to compare price elasticities for various goods

or for various factors of production because the'unit of analysis

is the same throughout. But comparison of a "wage/concentration"

elasticity with "wage/profit" or "wage/education" elasticities does

not have the same appeal because of the very different units used to

measure the exogenous factors. Comparing the price elasticities of

apples and oranges has a common sense interpretation, but not so for

a comparison of the earnings effect of years of education and after-

tax profits.

A not unrelated problem arises m the non-marginal nature of

variation in the exogenous factors used in this study. Infinitesimally

small differences in human capital or industry and occupation

characteristics do not accurately characterize changes in these

variables. Normally we are interested in the effect of an additional

year of schooling--or even the attainment of a diploma or degree- -

not the.impact, say, of a one percent increase in schooling past the

eighth grade. The same can be said for concentration and other

industry factors. For thistreason, Weiss, for instance, uses given

levels of unionization and concentration in evaluating his equations,it

not elasticities.1 In the final analysis, what we are after is a

measure of some range of earnings over so ratlike of its determinants.

Such a range can be estimated by measuring continuous variables

at arbitrary distances from their means and measuring dichotomous

1See Weiss, "Concentration and,Labor Earnings," op. cit.

243

226

variables at zero and one (e.g. no-training/training). One con-

venient method is to evaluate individual variables at ± one standard

deviation from their mean 1.7lues as we did in-the last chapter.

For a normal dis -fbution this yields a range over the middle 2/3 of

the observations. For other than normal distributions, the range

seldom includes less than 1/2 or more than 2/3 of all the observations,

making this measure variable, but bounded. Such a measure, of course,

does not cover the full range of a variable's distribution and there-

fore in most cases provides a somewhat conservative estimate of the

potential total impact of a given exogenous factor:2

In our desire

to err on the conservative side if necessary, this is a satisfactory

measure if only a single variable is to be evaludted.

But by its nature such a single variable measure cannot provide

an unbiased estimate of the impact of a combination of factors

analyzed ad conjunctum. For present purposes such a technique is

required for ultimately we wish to estimate the earnings impact of

employment in a given multivariate "economic environment"--defined by

a combination of an industry's concentration, profitability, and say,

capital-intensity or the combined effect of industry and occupational

segregation. JThe ad seriatum measure tends in almost all instances

to give pward biased estimate of the combined range and in fact

may result in evaluation of the regression at points well outside of

2Recall Chapter V fn. 3 for an extended discussion of this

evaluation technique.

244

227

the data's regime. It may happen, for instance, that within all of

the observations in a given occupation stratum, no single individual

can be found in an industry which is simultaneously lo greater on each

of the separate industry measures. In this case it is obviously

improper to evaluate the equation by summing the ±10 wage differentials.

To overcome'this deficiency a multivariate measure was devised

that accounts for the actual variation in the exogenous variables,

taken as a unit.3

Use of this measure normally prevents an estimate

of a wage differential larger than the data's full regime and virtually

always smaller than the ad scriatum estimate. Consequently it tends

to further restrict the measured wage range due to industry and

occupation variables--once more yielding a conservative estimate of-

these factors. Separate unit estimates were made for the stratification

and industry modules. In evaluating the equations for the impact of

. "complex crowding," the two-estimates were then added together.4

The Z* Measure

The ad conjunctum measure used in this part of the analysis

involves estimating the standard deviation of a linear combination of

the continuous variables in a given module using the regression

31 am indebted to Prof. Malcolm Cohen of the Institute of Laborand Industrial Relations, University of Michigan for suggesting this

measure to me.'

4This may lead to a slight upward bias infthese estimates for

precisely the same reason that we rejected the ad seriatum measure, but

the_opposite signs on the stratification and industry module variablesprecluded the use of a joint ad conjunctum technique.

245

228

coefficients as scalars., The standard deviation thus derived will

be known as Z*, not to be confused with z-transformations or other

statistical parameters. A Z* range is calculated for eacliindustry and

stratification module based on the regression equations reported in the

last chapter. The derivation of this multivariate measure is generally

straight-forward.

Let aibe the estimated regression coefficient where X

ijis

the jth observation on the ith continuous variable. is then the

jth linear combination of the Xivectors.

+ a. X = Zm ml , 1

+ . . . -17 ,aniXm2 = Z2

a1Xln

+ a2X2n

+ + amXmn

= Zn

or in vector notation:

a1X1+ a2X

2+ +am X

m= Z

From this set of linear combinations, the mean of Zj (=2) can be

calculated as well as its standard deviation Z*.

_2)2

N-1

The measure ±Z* then provides a direct reading of the range in the

exogeneous variable due Ito the combined variation id the Xi's. In the

246

229

present cases ±Z)g is the ad` conjunctum measure for the effect of the

stratification module (excluding the dichotomous variable, "union

mem:.,er") while +1* is an analogous measure for the industry module.

Intuitively, --ZS is the wage differential associated with an

industry=occupation "envirunment" which has "dne standard deviation".

less minority emp' yment. The estimate +Zt is the wage differential

associated with a .dssive economic environment" assessed on the

basis of such factors as concentration, after-tax profits, capital-

intensity or governmeto- "Nmand.

The superio' y of this unit measure over the ad serlatum

technique cin be demonstrated, first, by specific example and then

more generally. It will be shown that the ad Seriatum estimate is

always biased upward except in .he improbable case of perfect

positive pairilise correlation between-exogenous variables. The

.ollowin6 simple but generalizable two-variable two-obserVationexample

demonstrates the bias in the ad'seriatum measure and the corrected '

estimate generated by the 1* method.

Assume a regression has been generated for Y containing two

5observations and two dummy independent variables,'X1

and X2'

In

order to simplify the example, let the final regression have the

form: Y = .251*

, + .25X2 ;Pc. With this limited information we can1,72

.

. ,L.

. ,

Compare the ad eriatum (Z') and aLiEDILECL2m (Z*) evaluations oft

the X module under the assumption of a positive correlation between

X1and X2. 1, this case, the values of the evaluation estimates

will be identical (Z'=Z*).

_5Obviously such a regression coold not actually be generated

- 40* -

Frf

230

Ad Seriatum (Z') Ad Conjunctum (Z*)

X1

X2

ok

.25 + .2511/ 1

Z' = .25a + .250.X1

= .1761+ .1767

Z' = .3535

X1

X2

Z

' [(.25 x 0) + (.25 x 0)) = -0

[(.25 x 1) + (.25 x 1)) = .50

Z* = .3535

In the opposite case where Xi and X2 are negatively correlated, the

Z' and Z* evaluations are no longer equal, the former generating a

value 119 different from the case positive correlar, but the

latter equal to zero:

Ad Seriatum (Z') Ad Conjunctum*(Z*)

Ai X2 X1 X2

( )

0 [(.25 x 0) + (.25 x 1)] .25

.25 .25[(.25 x 1) + (.25 x 0)] = :25

1 0

L' = .250 + .250X1

X2

= .1767'± .1767

Z' = .3535

because of its singularity.

248

Z* = .0000

O

a

231

In this case the "standard deviation" of the X module as

measured by Z*,is zero because of the perfect offsetting impacts of

X1

and X2

(given identical reeiession coefficients). This is, of

course, the correct estimate of the differential in Y due to the

combined effect of the Xi, for any t'gain" due to having the

Characteristic X1

is simultaneously offset by an identical "loss" in

Y due to the absence of X2,and vice-4 la. For the analysis at hand

this would be similaf to2situatior, where all industries with

greater than average concentration had less than average profitability.?

Empirically the zero-order correlations for the industry and stratifi-

cation variables are usually positive but far from unity. Consequently

the Z* measure corrects for potential overestimates generated by the

ad seriatum technique.

A more general demonstration of the properties of the Z* measure

can be, provided, again using two variabThs and two observations for

expositional simplicity. What is to be proven-4s that:

(6.1)

(6.2)

lim Z* = Z'

pX1X2'

++1

lim Z* F 0

PX X 4-11 2

if a1

= a2and where pis a zero-order correlation

coefficient between independentvariables

4

249

.(6.3)

232

Define the ad seriatum measure in the usual fashion:

Z' = alaX + a20X2

1

and let Z*.be the standard deviation of the linear combination of

independent vectors (Z). The.derivation of Z* is straightforward.

(6.4) . a1X11

+ a2X12

= Z1

aX + a2X = Z1 21 2 22 2

From (6.4) the mean of Z (Z) equals:

a1(X

11+ X

2) + a2(X12 + X

22) Z

1+ Z

2= = Z

2 2.

Therefore,

(6.5) a1X1 + a2X2 =

-411.-041°

The standard deviation of Z follows directly btdefinition:

az = z* = ozi - E)2,(14 -31/2

Solving for Z* in terms of X1 and X2 can be done by first solving

for the squared deviations.

(6.7) Z1-Z=a1 X

11+ a2X12

1a2R2

= al(X.. )X1)

a2 X12- X2)

250

233

Analogously,

.

(6.8) Z2- Z = a

1(X

21- X

1) + a

2(X22

- R2)

Then squaring both sides of (6.7) and (6.8) gives

2 al(' 11-n 251

2 - .2(6.9) (Z1-2) X X1) + 2a

1a2(X

11-1

1)(X

12-2 ) + a2(X12-X2)

and 2)2 = a2 (X121-=1'2 + 2a a (X 5iX) ) 1221-. ) (X22 2' 2 22 2

a2fx.

1

2

0

,And summing the squared delPations

(6.10) (Zi-i)2 = al[(X11 -X1)2 (X21-5(1)2]

+ 2a1a2[(X1lky(X12-R2)(X21-5i1)(X22-312)]

+ a2[~(X22 7(2) 2]

Finally dividing both sides by N-1 (=1) gives the variance in Z.

2(6.11) Z*2 = a2a

2+ 2a

1a2cov (X

1X2) a

2

2aX1

1 . 2

Now for Z*2

to equal Z'2, titer

2 2(6.12) a

1aXi

+ 2a1a2cov(X

1X2) + a

2a2

(a + a2aX

)2

X2

1 X1 2

2 2 2= a

1

2aX

+ 2a 1a2aX aX + a2aX2

1 2

251

which simply reduces to

(6.13)

234

cov(X1X2) = a

X1aXi

Finally, dividing both sides by ax ax provides a proof of (6.1).

1 2

(6.14) cov(X1 X2)=p x = 1

lr.2 7m.X1

2

ax1aX

2

Q:E.D.

To prove (6.2) divide both sides of (6.11) by and set'a1 =a2..a.

This gives

(6.15) 2 2 2 2a a

X1

2a2cov(X1X2)

a aX2

a aXi X2 X1 X2 X1 X2 X1 X2

Then setting cov(X1 X2)/(ax1 ax2) = P

(646)a a

1

a2aX27,*

2 X2a

2

2

ax1aX

2 2

aX

= --1 and cancelling yields

Remultiplying both sides of (6.16) by ax1 ax2 yields:

252

235

(6.17) 2 2 2 2 2 2Z* = a 0 - 2a aX

X2+ a ax

1 2 2

= (ao - au )2

1X2

Finally, taking the square root of both sides leaves an expression .

for Z*

(6.18)

Thus when P

Z* = aoX1

- ao =ra(o' - aX

)X2

X1 . 2

Z* ...: 0 if either of two conditions holds:

(1) a1=a

2=a=0

or (2) aX

= aX' when a1=a

21 2

Q.E.D.

The first condition is trivial, showing only that the X module has no

impact on Y when the regression coefficients on Xi

are insignificant.

Condition (2) is more substhitive, demonstrating that the impact of a

giVen module is zero when there is identical variance in all of the

. exogenous factors and the variables are inverse correlates of each

other. Thus the multivariate measure has the property of ranging from

zero to Z' as the correlation between paired explanatory variables

runs from negative to. positive one'. This is, of,course, a desirable

property for such a statistic.

253

236

The Results

In the actual estimates that follow, a Z* is calculated for

the industry and stratification modules wherever there are two or

more, continuous variables in a given module. Otherwise the equivalent

ad seriafum measure is used. Where the dichotomous variable, "union

member" is significant in a regression, it is evaluated at zero and

one and added linearly to the,estimate of Z*. Consistent with the

rationale for "simple" and "complex" crowding, the regression

equations are evaluated at (a) the mean for all variables (W), (b)0

then'at the mean for ail of the variables excluding the_ stratification

factors which are evaluated at ( ±Z * ±UN), and finally (c) at the mean

for all of the variables excluding those in the stratification and

industry modules both okwhich are evaluated according to the Z*

formula. This final statistic then measures the overall range in

earnings for a human capital constant population evaluatedin terms _

of'

±Z*S

±UN, and ±E*. All of these range or interval estimates are

based on the regression equations recorded in Chapter V. The tabular

results that follow report hourly and annual earnings intervals as

well as associated percentage differentials.6,7

Each race-sex group

6In terms of annual earnings, full -time full-year employmefit

is-assumed to be 52 weeks x 40 hours per week = 2080 hours/year.

7The two percentage earnings intervals are calculated in the

following way:

(-Z* +UN -IV) - (+X* --,UN'-Z*)

(1) "COMPLEX"(+X*

S-UN -E*)

2371

is reported separately and followed by the results for the labor force

as a whole.

White Males - As expected, the narrowest wage differentials due

toeXIsting-variation in non-human capital factors are found among

white men. Nevertheless these differentials are far from

inconsequential particularly in the lowest skill strata. '(See Table

6.1) The results for occupation group.1-3, for instance, establish a

perfect example of the "simple crowding" phenomenon. Holding human

capital fixed, a full $1.00 an hour wage differential is found based

on an evaluation of the STRAT factors alone. On an annual basis this

amounts to an almost $2100 interval around a mean of $5637. The worker

in a "permissive economic environment" (based on union membership

and the degree of minority crowding) can expect on average Ito earn

nearly 1-1/2 times (146%) the earnings of a similarly skilled non-union

worker in a minority-crowded industry and over 17 percent mere than

' the average wage in this stratum. In this particular case the

comparison is between a union,worker in an industry with 14 percent

minority employment and an equally skilled but unorganized employee ,

in an incluse?), which has over 46 percent of its labor force compo'sed

of white women and blacks of both sexes. In other strata the differ-

ential is by no means cc large, but still exists.

(-zg +um) - (4.4 -UN)

(2)

(-1-z* -UN)

255

"SIMPLE"

\

..

238

TABLE 6.1

WAGE INTERVALS DUE TO STRATIFICATION ANDINDUSTRY FACTORS, BY OCCUPATION STRATUM

WHITE MALES

OccupationStratum

1-3

+ UN + Z*

SUN

+I* - UN

+Z* - UN - Z*

5

-ZS + UN + Z*I

-Z*S+ UN

W

+Z*S

- UN

+Z*S

- UN - Z*I

6-9

-ZS + UN + ZI

-ZS + UN

W

+Z*S- UN

+Z*S- UN - ZI

Deviations from P

Total EarningsIntervala

W Annual W % I

$3.18 $6614 17.34% $1.00 45.87%

3.18 6614 17.34 1.00 45.87

2.71 5637

2.18 4534 -19.55

2.18 4534 -19.55

3.17 6594 10.45 .61 23.82

2.98 6198 3.83 .23 8.36

2.87 5970

2.75 5720 -3.83

2.56 5325 -10.45

3.41 7093 15.20 1.00 41.49

3.41 7093 15.20 1.00 41.49

2.96 6157

2.41' 5013 -18.58

2.41 5013 -18.58

56

239

TABLE 6.1 (Continued)

OccupationStratum

Deviations' from 171

Total EarningsIntervala

W Annual W

12-14

-Z* + UN + Z* $3.84 $7987 9.71% $ .85 28.42%

-2* + UN 3.75 7800 7.14 .44 13.29

3.50 7280 --

+Z*S- UN 3.31 6885 -5.42

+Z* - UN - Z* 2.99 6219 -14.57

15-17

-Z* + UN + Z* 5.11 10628 5.82 .56 12.40

-Z* + UN 5.11 10628 5.82 .56 12.40

P 4.83 10046 --

+2* - UN 4.54 9443 -5.85

+Z* - UN - Z* 4.54 9443 -5.85

All Strata

3.78 7862 10.52 .68 21.93Z-ZS + Ur +I

-ZS + UN 3.52 7322 2.92 .16 4.76

W 3.42 7114 --

+Z*S- UN 3.36 6989' -1.75

+Z*S- UN - ZI 3.10 6448 -9.35

I

The first row of statistics reports the interval between-Z* + UN + Z* and +Z* - UN - Z*

The second row reports the interval between -ZS + UN and +qc - UN

257

240

Occupation stratum 5, as one may recall, is comprised mostly of

operative and kindred workers. The evidence clearly indicates that

there is less variation in wages due to industry factors in these

fairly homogeneous occupations. Yet differences in .the extent of

minority employment by industry and concentration account fora $.61

earnings wedge between equivalent workers. On an annual basis this

amounts to a $1270 earnings gap or 24 percent. If we were to disregard

differences in industry demand characteristics and only evaluate the

regression for variance in the stratification module, the total wage

range would be only $.23 or 8.36'percent. Much of the tiotil wage

differential is consequently explained by differences in industrial

concentration given initial labor supply restrictions.

In occupation stratum 6-9, composed of many of the skilled

trades, union membership plays the critical role in the distribution

of earnings. Union membership alone is worth $.76 an hour (see

Chapter V) out of a total wage differential of $1.00, the remaining

gap due to the fact that apparently some white men are "trapped" in

occupations crowded with black male workers. The $1.00 an hour

amounts to a 41.5 percent earnings differential between workers of

apparently equal endogenous productivity. The difference on an annual

basis is $7093 vs. $5013.

A significant wage differential even prevails among white male

workers in the relatively highly skilled occupation stratum 12-14.

Here there is a $.44 earnings gap between workers who differ by ±Zt and

union affiliation and an additional $.41 due to differences in the

/4.58

241

industry module. Summed together this drives a 28 percent wedge

between the annual earnings in a "permissive" vs. "repressive"

economic environment.

Only for the very most skilled white male professional workers

is the differential relatively unimportant. Here non-human capital

factors are Iesponsible for nomore than a 12.5 percent wage gap

between similarly qualified workers and the full extent of this range

is apparently related solely to differences in industry profitability.

When we turn to evaluate the white male equation across all

occupation strata, 'thus accounting for the full effect of human capital,

we again find a relatively large wage differential due to stratification

and industry factors, particularly the latter. Stratification factors

(after controlling for the interaction between union membership and

concentration) produce only a 4.76 percent wage differential. Once

the Z* is added, however, the total earnings gap rises to $.68 or

nearly 22 percent. On an annual basis this amounts to a more than

$1400 differential, with earnings ranging from $3:10 an hour to $3.78.

While these industry and stratification associated wage differentials

are much smaller than for each of the minority groups, they are by

no means insignificant and certainly too large to ignore. The major

unanswered question is how to explain them.

Where much of the earnings differentials between race-sex

groups can be charged to discrimination in its many forms, this

explanation is mostly useless for the dominant white male group.

However, a number of possible alternative explanations can be4..

759

242

ascertained. One hypothesis consistent with radical stratification

theory maintains that wage differences among similarly qualified

white men are due to unspecified variation in the workers' social

class origins. Accordingly, higher wage workers have benefitted from

being nurtured in an environment of financially and socially well-to-do

families. Unfortunately we have not been able to control for this

factor due data limitations. Ultimately the social class hypothesis

may explain some of the wage difference associated with industry

and stratification factors, but at this point we have no proof.8

Another explanation might lie in compensatory wage payments

-which do not show 'up in the analysis of earnings or in fringe benefits

that are inversely correlated with straight-time hourly wages. What

evidence we have,on compensatory payments seems to indicate just the

opposite however. The phAical demands variable in the white male

cross strata equation is significant but negative. Little hard

evidence exists on the fringe benefit question, but casual observation

seems to indicate a probable positive correlation between wages and

8What evidence does exist on thisisubject tends to deny the .

importance of social class as a determinant-of the variance in income.In his study of Inequality, Christopher Jencks concludes that in factmost of the variation in men's incomes appears to be stochastic:

"Neither family background, cognitive skill, educationalattainment, nor occupational status explains much of the vari-ation in men's incomes. Indeed, when we compare men who areidentical in all these respects, we find only 12 to 15 percentless inequality than among random individuals."

A

Christopher Jencks, Inequality: A Reassessment of the Effect'of Familyand Schooling in America (New York: Harper & Row, 1972), p. 226.

"i%

ti

243'

non-wage supplements.

A more plausible hypotA'sis relies on,the existence of

widespread imperfections in information about job opportunitiO. This I

of course makes, a good of sense at least as an explanation of

short-run wlge differences. Such imperfections_could-well-explain--- --

wage intervals of the magnitude found in the higher skill categories.

Larger more permant_nt differentials, it would seem, require a more

cothplex hypothesis.

One such possible hypothesis can be derived from a synthesis of

thedries based bn tne work of Thurow and. Lucas9

(the "job competition"

thesis), 'Decker10

and Oi11

(the concept of labor as a "quasi- fixed"4

factor) and the institutionalists (the importance of "lock-i n" effects

in the supply of labor). ACcording to the job competition thesis,

,individuals compete for jobs based on their background characteristics,

not in terms of crape demands as standard neoclassical theofy suggests.

One can imagine a queue of jobs defined by a set of characteristics

the hourly, wage rate being o..e of the defining ?arameters.12

\9Lepler C. Thurow and Robert E.B. Lucas, The AmericariDistribution

,1 Income: A Structural\Problem, A Study Prepared for, the JointEconomic towmittee of the U.S. Congress (Washington, D.C.: U.S.Government Printing Office, March 171 1972), esp. pp. 19-39.

10Gary Becker, Human Capital, Chapter 11, CD. Ci.t.

11WQter0i, "Labor as a Quasi-Fixed Factor," Jce,Jrnal of'Political

. ,

Economics, December-1962.

12One very difficult q uestion is left unanswered by the job

comp etition model: what determtnes the distributiod,of wages in the

first place? If labor supply and demand facto:s are so weak as toloavl the wage indeterminate, what other factors define the actual wage

261Mln

Workers compete for these job/w43e slots by presenting themselves in

the job market to pote:ttial employers. Firms then choose employees

on the basis of expected training costs (given their background

characteristics), hiring first those with the lix..est expected employment

cost and then moving down the queue to higher cost labor if detaind

warrants.

If we apply this model over the business cycle, we can genetate

a pattern to that at any given point.in time workers of identical

endogenous productivity will be found in different job slots and thus

earn various wage rates. This will occur as\a worker who enters .he

job market during a period of tight demand will have a greatei

probabi/ity of finding a higher paying job while the worker who oins

the market in a contractionary period may have to acceptalwr

paying job for the same amount of search effort. If search c sts were

low, the fixed cost of hiring and training labor wen minimal, and

there were no substantial "lock-in" effects, earnings differentials'

would only be temporary for lower wage workers would con nually reenter

the job 'market in an attr-ot to gain employment in the gher wage job

slots consistent with their endogenous-productivity. A strong tendency

paid on a given job? One posstble answer is that s-pply and demand

are responsible for getting a wage,range"for every "job" but that

custom and inertia--as well as institutional factors including union

pressure - -are responsible for setting and holding the wage distribution

as it -is. Once established the pattern of wages changes only slowly

in response to real changes in supply and demand'. For the most part

the wage distribution is never in equilibrium accounting for a good

deal of-structural unemployment in all labor markets.

\245

toward equal returns for identical personal characteristics would

be-the consequence.

.In fact, however, labor is usually a "quasi-fixed" factor,

search *costs are often substantial, and "lock-in" effects are

extensive. Specific training costs will often be shared by both the

worker and the firm (with the shares depending on expected turnover

and quit rates).13

Once workers have invested in specific training

in a particular slot, their marginal products and therefore their

wages are presumably higher than in alternative employment. Consequently

a worker will tend to remain in a job for which he has already paid

for training rather than quit to begin a new job,at a lower wage rate

in hopes of working up to a higher one. Employers too will be

reluctant to dismiss already trained employees so as to hire replace-.

ments even if the potential recruits embody superior background

characteristics. Thus where labor has a high degree of "fixity," to

use ()its term, there will be a tendency for workers to stay where

they are (and employers to keep them) even in the face of fairly ,

substantial differences in hourly rates. This, of course, is fully

consistent with individual utility functions which posit that workers

attempt to maximize the expected value of lifetime income rather than

simply maximize their wage.

The foregoing eclectic theory, is obviously suggestive for the

more skilled workforce, those in our sample with high SVP levels for

13See Becker, Human Capital,,alcit., pp. 21-22.

Z63

246

instance. But the largest wage differentials, due to other than

human capital factors are found among the least Skilled workers,

presumably those. with. a low degree of "fixity." For them the

"quasi-fixed" factor theory does not directly apply, but an inst

tutional variant along the same theme does. Specific training/4d

hiring costs produce one form of "lock -in" effect, where the/More

common mechanisms are seniority privileges and non-vested p nsions,

both of which apply to the full occupation spectrum, the owest skill

strata included. In attempting to maximize expected lifetime incom1'

.a worker with many years of seniority and associated pensi n rights

will not move to a job.with a higher hourly wage rate if this means

sacrificing the employment security which goes along with seniority,.

(particularly in unionized firms) and the surrender of expected

retirement income. In this case fairly large wage differentials will

persist over time once the differentials exiit at all.

-Unfortunatelywe do not have any data tq test this',Ilypothesis,

but it seems a likely candidate to explain the substantial and

probably persistent wage differences found among all but the most

gilled white male workers. "Entrapment" through fixed training costs,

imperfections in information, and non-vested seniority and pension

privileges may very well be responsible for driving a wedge of as much

as $2100 in annual earnings between white male workers who have

substantially the same huMan capital attributes.

Black Males - Once we leave,the realm of white male workers,

the,impact of industry and stratification factors becomes much more

Z,64

247

significant. This can readily be seen in an evaluation of the black

male regressions. (See Table 6.2) In virtually every one of these,

there is extensive evidence of "complex crowding" with union member-

ship playing a consistently effective role in every stratum. The

percentage earnings gap is as high as 75 percent (0cc Stratum 1-3)

and the annual dollar difference, according to our evaluation

echni5ile; teaches almost $2800 (0cc Stratum 12-14).

Union-nlembership and Industry segregation are responsible for a

5 percent differential among black men in the lowest skilled occupa-

tion category. Adding the combined effect of di"ferences in concentra-

tion and government demand raises the total differential to 75.8 percent

or a $1.20 an hour range around a mean of only $2.17. The stratification

and industry modules apparently contribute about equal weight to the

overall wage gap. In occupation group 5 composed predominantly of

4 operatives and. janitors and sextons, the total earnings differential

is of almost identical magnitude (74.9%), but nearly wo-thir of

the total is due to stratification factors--mainly unio hership--

whil:e the remainder is due to the single industry factor, concentra-

, 4

tion (see Chapter V). This is in sharp contrast to the white male

regression for this stratum where we found only a small earnings

differential (23.8%). Of this only a quarter, was due to stratification

factors and union membership apparently played no role at all. The

rest of the relatively small $.61 differential was due to differences

in concentration..

ZE';5

248

TABLE 6.2

WAGE INTERVALS DUE TO STRATIFICATION ANDINDUSTRY FACTORS, BY OCCUPATION STRATUM

BLACK MALES

Occupation

"Stratum

Deviations from i4

Total EarningsInterval

W Annual W

1-3

-ZS + UN + ZI $2.78 $5782 28.11%"

$1.20 75.80%

-ZS + UN 2.51 5221 15.66 .65 34.94

W 2.17 4514°

+Z* - UN 1.86 3869 -14.28

+Z* - UN - Z* 1.58. 3286 -27.13

5

-ZS + UN + ZI 2.99 6219 25.10 1.28 74.85

-ZS + UN 2.75 5720 15.06 .80' 41.02

W 2.39 4971 --

+Z*S- UN 1.95 4056 -18.41

+Z*S- UN - ZI 1.71 3557 -28.45

6-9

-ZS + UN + ZI 2.87 5970 21.73 .95 49.23

-ZS + UN 2.70 5616 14.40 .60 28.57

ii 2.36 4909

+Z*S- UN 2.10 4368 -11.01

+zg - UN 7 Zt 1.92 3994 -18.43

2,66

1

249

TABLE 6.2 (Continued)

OccupationStratum

Total EarningsDeviazions from P Interval

W Annual 'W

.12-14

-ZS + UN + ZI $3.32 $6906 28.64% $1.33 66.83%

-ZS + UN 3.05 6344 17.76 .79 34.95

W 2.59 5387 7-

r $

-4-ZS - UN 2.26 4701 .712.74

4Z*S- UN - ZI 1.99 4139 -23.16

15-17

-Z* + UN + Z*

-Z* + UN /1/

SAMPLE SIZE TOO SMALL FOR SIGNIFICANT RESULTS

+Z* - UN

4Z* - UN:- Z*

All Strata .

-ZS + UN + ZI 3.01 6261 27.00 1.15 61.82

-ZS + UN 2.82 5866 18.98 .77 37.56

W 2.37 4930

+Z* - UN 2.05 4264 -13.50

+Z* - UN - Z* 1.86 3869 -21.51

250

The total wage gap in occupation stratum 6-9 is smaller than

in the other strata, a perplexing result at first glance. The full

interval is 49.2 percent, not much greater than the differential for

white men although still equiiialent to almost $2000 on an annual basis.

The relatively lower earnings gap is apparently related to weaker

effects of both unionization and concentration but even more so to the

virtual absence of any significant segregation factor. The perplexi

result is made comprehensible-once we recall that when segregation

in a particular race-sex group is overwhelming, the true earnings

differential may be empirically undetectable. The differential can

only be uncovered by evaluating the pooled race-sex regressions.

M6ving to the higher skilled occupation stratum 12-14, we find

the percentage earnings range among black men to be more than double

that of their white male counterparts and the dollar gap reaches a

maximum for any group in any'stratum ($1.33 an hour). About half

the total differential is associated with the stratification module

while the remaining half is due to differences in concentration. BaSed

on the evaluation procedure, estimated hourly wages for this group

span the interval $1.99 to $3.32. Unfortunately the data sample

does not provide enough observations on professional black men to

test whether the earnings differential substantially declines as for

white men.

In turning to an examination of the cross occupation regression,cx

one is immediately struck by the fact that the total percentage

earnings differential is almost three times that for white men. The

2 68

251

estimated range runs from $1.86 to $3:01'an hour compared with an

estimated range of $3.10 to $3.78 for white males. Of the full

$1.15 an hour wage spread due to stratification and industry factors,

$.77 is due to the "supply side" with the remaining amount the

effect of a linear combination of after-tax profits, capital/labor

ratios, and government demand. The average black man in the full -time

SE0 sample earned $4930 on an annual basis, but given "average" human

capital characteristics, the same worker could earn anywhere from an

estimated $3869 to $6261 depending on how fortunate he was in finding

employment in an industry characterized by a "permissive economic

environment."

Much of this massive earnings differential may be explained by

the same factors as we hypothesized for white men: compensating non-

wage supplements, imperfections in labor market information, and

lock-in or entrapmbnt effects. But in addition to these there is

considerable evidence of specifically race-linked segregation. The

estimated,STRAT module induced wage interval for the white male

pooled regression is only $.16 an hour compared with the $.77 range

estimated for black men. Part of this large difference is due to the

much stronger impact of union membership on wage differentials while

the remaining is due to the greater impact of the industry segregation

factor %MININT.

White Females - The tale told in the evaluation of the white

femae.regression results is a similar one, but even more difficult

to uncover because of a much greater degree of occupational

269

252

segregation. The estimated percentage differentials generally lie

between those of comparable white and black men. (See Table 6.3)

In the lowest skilled-category, only the stratification module is

significant but union membership as well as a linear combination of

both occupation and industry segregation provide a 42 percent earnings

interval with a dollar value of $.62 around a mean of $1.74. The

total differential in occupation stratum 5 is somewhat larger (54.6%),

but here the range seems to be better explained by differences in

industry characteristics wich the stratification module contributing

only $.24 to a total $.87 differential. Union membership is the only

significant STRAT factor in this regression.

Again as in the black male results, the earnings gap in occupation

group 6-9 is lower than in any other stratum (with the exception of

the professional group). The total gap is $.56 or 35.4 percent. A

smaller coefficient on the union memberhip parameter seems to suggest

the reason for this relatively narrow range in wages. But it is the

smaller variance in this factor due to the underlying high degree of

industry segregation that really explains this result.

This same.effect is nowhere more evident than in the top two

occupation categories where in both cases the regression coefficients

in the stratification module are insignificant thus yielding a manifest

earnings range of zero associated with these factors. The nearly 50

percent total wage differential in occupation stratum 12-14 appears

to be solely due to an ad conjunctum analysis of after-tax profits

and capital/labor ratios while the smaller 28 percent differential in

270

253

TABLE 6.3

WAGE INTERVALS DUE TO STRATIFICATION ANDINDUSTRY FACTORS, BY OCCUPATION STRATUM

OccupationStratum

1-3

-ZS + UN + ZI

-ZS + UN

ci

+Z*S- UN

+Z*S- UN - Z*

I

5

-Z* + UN + Z*

-ZS + UN

W

+Z*S

- UN

+Z*S

- UN - Z*I

-61-9

- ZS + UN + Z*I

- ZS + UN

i:i

+Z*S- UN

+Z*S- UN - Z*

I

WHITE FEMALES

Deviations from W

.Total Earnings

Interval

W Annual W % $

$2.11 $4389 21.222 $ .62 41.54%

2.11 4389 21.22 .62 41.54

1.74 3619

1.44 2995 -14.36

1.44 2995 -14.36

2.47 5138 22.69 .87 54.63

2.15 4472 6.96 .24 12.56

2.01 4181 --

1.91 3973 -5.00

1.59 3307 -20.65

2.14 4451 16.30 .56 35.44

2.03 4222 10.32 .34 20.11

14 3827

1.69 3515 -8.15

1.58 3286 -14.13

254

TABLE 6.3 (Continued)

OccupationStratum

Deviations from 14

Total EarningsInterval

W Annual W Z $

12-14

-ZS + UN + Z*I

$2.87 $5970 22.07% $ 05 49.63%

-ZS + UN 2.36 4909

W 2.36 4909

+Z* - UN 2.36 4909

+Z* - UN -.Z* 1.92 3994 -18.42

15-17

-Z* + UN + Z* 3.18 6614 12.35 .70 28.22i

-ZS + UN 2.83 5886 -- *MO.

II 2.83 5886 --

4Z*S

- UN 2.83 5886

41 - UN - ZI. 2.48 5158 -12.38

All Strata

-ZS + UN + Z*I

2.59 5387 26.01 .96 58.22

-.ZS + UN 2.36 4909 15.12 .49 26.20

W 7 2.05 4264 --

+Z*S

- UN 1.87 3890 -8.78

4Z*S

- UN - ZtI

1.63 3390 -20.36

272

255

the professionals category appears purely as the result of variance

in concentration. Labor supply imperfections not specified in the

regressions, such as those used to explain the white male wage

differential, are probably responsible for permitting the labor demand

variables to have such a significant impact on the estimated earnings

gap. Again it should be noted that the smallest wage interval is

among the professional class while large differentials permeate the

rest of the occupation strata.

In turning to the cross occupation estimates, we find a dotal

wage differential (in percentage terms) not significantly di ferent

from that of black men. In this case the total differenti is equal

to $.96 an hour or 58.2 percent.. A little less than half of this

differential is associated with the STRAT module while the remainder

is due, again as with black men, to a linear combination of after-

tax profits, eqltal/labor ratios, and government dematf1. The

overall wage interval runs from $1.63, just barely above the 1067

prevailing minimum wage, to a high of $2.59 an hour for women who

gain access to industries or occupations characterized by a

"permissive economic environment." In explaining these intra-group

diffeentials we might rely on the same hypotheses we posited for

white and black men and add the theory concerning different utility

,,functions for women in different objective situations that we outlined

in Chapter V.14

14See Chapter V, fn. 35, p. 205.

256

Black Females - The extraordinarily large wage differentials

found for black men are repeatedlrfob-latkwomen, u_th the exception

that in occupation stratum 6-9 the earnings gap is even larger.

(See Table 6.4) Being in a permissive economic environment can mean

as much as $1.17 improvement over those who are not as fortunate,

but given the very narrow range of opportunities for black women,

even a "permissive economic environment" leaves virtually all of the

' workers in the first three occupation strata with estimated annual

earnings below $5,000./ In each of these cases, the largest part of

the overall differential is due to stratification factors with union

membership significant in every regression.

/

An evaluation of the pooled strata equatim turna_pufto yield

an ear..ings range which is almost identical in percentage terms to

those found for the other two minority groups, although in this case

a greater pibpoktion of the total differential is associated with the

stratification factors. Only concentration,is significant in the

industry module and at best variance in this measure adds $.15 to

the $.82 differential. ,Table 6.5 demonstrates the near identical

percentage differentials for the three minority groups. This striking

similarity in the overall earnings differential is in sharp contrast

to the much smaller interval associated with differences in industry

and stratification factors for white men. Clearly the minority

guitips have something in common which they do not share with the

-dominant group in the labor force and it is far from their advantage.

274

257

TABLE 6.4

WAGE INTERVALS DUE TO STRATriICATION ANDINDUSTRY FACTORS, BY OCCUPATION STRATUM

. BLACK FEMALES'

OccupationStratum

\Deviations from W

Total Earnings'Intervals

W. Annual W

1-3

$1.79,

1.71

1.36

1.13

1.04

, 2.21

1.86

1.53

1.39

2.36

2.17

1.72

. 1.37

1.19

$3723

3557

2829

2350

2163

490c

4597

3869

3182

2891

4909

4514

3578

2850

2475

31.95%

25.73

-16.91

-23.50

26.88

18.81

--

-17.74

--25.26

37.20

26.16

--

-20.34

-30.81

$ .75

.58

.97

.68

1.17

.80

72.47%

51.32

69.78

44.44

98.31_

58.39

\\N-...

-2* + UN + Z*

-2* + UN

W

42*S

- UN

+Z* - UN - ZIS -

5

-Z* +'UN + Z1S L

-q + UN

W

4Z*S- UN

42* - UN - Z*I

6-9

-ZS + UN + 2*I

-ZS + UN

'T/

+Z* - UNS

4-2*S- Ui- Z*

I

275

258

TABLE 6.4 (Continued)

Total Earnings

Occupation Deviations from W Intervals

Stratum W Annual W

12-14

-2* + UN +Z*

-Z*S+ UN

SAMPLE, SIZE TOO SMALL FOR SIGNIFICANT RESULTS'

+Z* - UN

+Z* - UN - Z*\ I

15-17

-Z* + UN + Z*

-Z* + UN

SAMPLE SIZE TOO SMALL FOR SIGNIFICANT RESULTS

+Z* - UN

+Z* - UN - Z*

All Strata

-Z + UN + 74 $2.12 $4410 27.71% $ .82 63.07%

-Z*S+ UN 2.05 4264 23.49 .67 48.55

,.'

W 1.66 3453

+Z*S- UN 1:38 2870

+Z* - UN - Z* 1.30

-16.86

2704 -21.68

Z76

259

TABLE 6.5

POOLED OCCUPATION REGRrSSION WAGEINTERVAL ESTIMATES

or

Race-Sex GroupDo ar

Dif rential

PercentageDifferential

Black Males $1.15 61.82%

'White Females 96' 58.22

Black Females .62 63.07

White Males .68 21.93

f

Cross Race-Sex - The individual race-sex equations mask the

of "crowding" as the extent of segregation rises beyond some

point\ Nowhere is this more true than among higher-skilled white

females where occupational segregation is so extensive that the measured

effect of the stratification module is zero. For this reason the

pooled race-sex equations must be evaluated to correctly estimate

the impact of the industry and stratification factors. The results

/con;:irm a significant earnings effect in every occupation stratum ana

for the labor force,ias a whOle. (See Table 6.6)

This effect .is by far the greatest in the low skilled occupa-

tions. The total estimated earnings range in occupation stratum 1-3

is a startling $3350 around an annual full-time mean of $4722.

Workers in a "permissive economic environment" earn 35.percent more

tnan the average wage for this group and more than twice (110%) the

wage earned%by those in overcrowded unorganized competitive industries.P

Minority segregation by industry and occupation, com)ined with union

260

TABLE 6.6

WAGE INTERVALS DUE TO STRATIFICATION ANDINDUSTRY FACTORS, BY OCCUPATION STRATUM

ALL RACE-SEX GROUPS

'OccupationStratum

Deviations from WTotal EarningsInterval

W Annual W

1-3

-ZS + UN + ZI $3.07 $6386 35.24% $1.611 110.27%

-ZS + UN 2.97 6178 30.83 1.41 90.38

T',/ 2.27 4722

+Z*S- TN .,

1.56 3245 -31.27

+Z*S- UN - ZI 1.46 ' 3037 -35.68

5.

+ UN F Z* 3.08 .640 24.19 1.21 64.70

-Z* + UN '2.85 5928 14.91 .75 35.71

2.48 5158

+Z* - UN 2.10 4368 -0.32 /

+Z* - UN Z* 1.87 3890 -24.59

6-9

-ZS + UN + Z*I ..

3.36 6988 29.23 1.54 84.61

-Zt + UN0

'1.36- 6988 29.23 1.54 84.61

1,71 /.6C 5408

+Z*S

-- UN 1.82 3786 -30.00

+Z*S- UN 2*

I1.82. 3786 1*-- -30.00 ..

261

TABLE 6.6 (Continued)

OccupationStratum

,Deviations from T.1

Total EarningsInterval

W Antival W

12-14

$3.90

\ 3.58

3.29

3.08

2.77

5.35

5.01

4.63

4.2

'3.92

3:67

3.42

2.96

2.56

2.30

:

''

$8112

7446

6843

6406

5762

11128

10420

.9.630-,

8840

- 8154

7634

7114

6157

5325.

4784

18.54%

.8.81

°

-6.38

-15.80

15:55

8.20

-- -

-8.20

-15.1 55

23.98

15.54

-13.51

-22.29

$1.13

.50

1.43

.76 .

1.37

.86'

:

40.79%

16.23

36.47

17.88, ...

59.56

33.59

-Z* + UN + Z*

-ZS + UN

W .v.

+Z* - UN ,

S

+2*S

- UN - ZtL

15-17

-ZS + UN + Z*1

-ZS UNS__W

+Z* - UNS

+Z*S- UN - ZI

All Strata

-ZS UN + Z*S I

-ZS + UN

itt

+Z* UN

+Zt - UN - Zt

Z79

262.

membership, is responsible for a $1.41 earnings differential while

concentration adds another twenty cents to the overall range. In

annual terms the wage interval runs from $3037 to $6386 with human

Capital evaluated at the occ group means. overall 110.27 percent`

wage interval compares with 72-76 percent intervals for black males

and females and 42-46 percent for white men and women suggesting a

.strong racial and sexual component in the industry and occupation

distribution of low-skilled workers.

In cccupation stratum 5 this "discrimination" component appears

less pronounced as the pooled percentage earnings differential falls

within the range of the separate estimates for each race-sex group.

Overall there is a $1.21 wage interval around a mean of 4.48.' A

little more than half (35.7%) of the total interval (64.7%) is produced

by the STRAT module while the remaining is due to a linear combination

of concentration, after-tax profits, and capital/labor ratios. On an

annual basis the estimated interval is more than $2500 running from

$3890 to $6406.

Turning to occupation stratum 6-9 we once again find an

indication of the massive effect of industry and occupational discrimi-

nation. In none of the individual race-sex equations were any of the

industry and occupation segregation variables significant (with the

exception of union membership). But in the pooled regression three

of these factors are significant and powerful. Analyzed ad conjunctum,

%HININD, %MINOCC, and %BMOCC plus union membership are resnonsAde

for an 85 percent earnings differential. Of this total range,

263

unionization is responsible for a little less than half ($.67) while

the other three crowding variables make up the remainder of the $1.54

interval. After controlling for these factors, differences in

industry structure have no additional effect on the wage range

suggesting "simple" but substantial crowding.

In the two higher skilled strata as well there is evidence of

sizable wage differences associated with the industry and stratification

factors. There is a $1.13 wage gap (41%)'in occupation group 12-14

with an interval of $.50 associated with union membership combined

with a Zoc evaluation of MININD and %MINOCC. The remaining $.63 is

due to a linear combination of two industry variables: concentration

and after-tax profits. Even among professionals there is a 36.5

percent differential or an almost $3,000 annual salary interval after

'controlling for human capital characteristics.15 This is primarily

due to the sex-liAked segmentation of the professional labor market.

About half of the interval in this stratum is due to the single

stratification variable %FMOCC while the remaining amount is associated

once arain with a linear combination Of concentration and after-tax

profit:4.

15As has been the case throughout, we have evaluated these

equations assuming that SVP is a true human ,..apital ce;oponent, not a

function of Industry'or occupation segregation. Of course if we were

to.interpret SVP as a stratification variable--for which there is a

good deal of ity.tification--the wage interval would be muct larger

in a number of these equations including tha present one. ,ater when

we evaluate the effect of human capital, we shall assigrPthe whole

weight of the SVP factor to this module, surely an overestimate of

the pure human capital effect,

264

Finally we come to the "grand pooled" regression for the

whole labor force. Here we find for a human capital constant popula-

tion a total range of $1.37 an hour or $2,850 a year on a full-time

basis. This amounts to a 60 percent earnings differential between

workers in a!"permissive economic cnviponment" and those, who for'one

reason q another, are consigned to industries which are on the

"periphery" of the American industrial structure--industries which are

non-unionized; impacted with minority groups, low.profit, labor

intensive, competitive, and lacking support in the form of go ernment

contracts.16 A little more than half (33.6%) of the total interval

is associated with labor supply restrictions *bile the rest is due

to differences in industrial characteristics.17

Such large differences

16See Robert Averitt, The Dual Economy: The Dynamics of American

Industry' Structure (New York: W.W. Norton & Co., 1968), esp. Ch. 5.

17We should emphaS'ize again that these are conservative estinates

b'cause of our evaluation technique. If we were to estimate the wage

interval over the total range of the exogenous variables rather than

at ±lo around their' means, or if we were to use an ad seriatum measure.

of the interval, we would find a much larger earnings range due to

the'industry and stratification factors. Evaluating the "grand'pooled"

regression ad :seriatum rather than ad conjunctum increases the total

wage interval co $1.71 and the percentage differential to 83.5 percent.

Instead of an annual income spread estimated at ;72,850, the ad seriatum

interval is $3,550, twenty-five percent larger. The correlation

matrices for 'lie relevant varlet-les indicate the reason for the lower

ad sonhInctum estimate.The zero-order correlation between %MIWIND and %MINOCC in the

stratification module is .3887. Tne industry module correlation matrix

has the follouing values:

Concentrtn. At,:TX Pr. K/L Ratio Gov't Demand

ConcentraticnAfter-tax profitsHaGov't Demand

1.0000 .33201.0000

, .4531-.08721.0000

.0997-.0092.0533

1.0000

mmlim

due to factors other than measured human capital surely calls into

question Leonard Weiss's conclusion--and the assumption of most human

capital theorists- -that "The general picture is one of fairly

efficiently working labor markets, even where substantial monopoly may

exist."18

What we have found in this extensive analysis is significant

evidence of widespread mismatching between endogenous productivities

and marginal products. Workers with substantially the same human

capital attributes earn substantially different wages, much of this

apparently related to industry and occupation "crowding" with

variations in industrial structure and performance adding to the

overall Wage diSpersion. The personal earnings distribution, we have

shown, is to a far-reaching extent a function of institutional

factors well beyond the purview, let alone control, of the individual

worker.

The Relative Impact of Human Capital andNon-human Capital Factors

Before bringing this analysis to a close, there is one

additional question that warrants our attention. We have estimated

the earnings differentials associated with industry and stratification

factors, but not those which are due to variation in human capital.

How large are these in absolute terms and relat::.ve to the size of the

Z*Sand Z* intervals?

To evaluate the human capital variables, we have resorted to an

ad seriatup measure so as to avoid as much as possible the potential

18Leonard Weiss, 2p. cit., p. 116.

',Z83

266

error of underestimating the full impact of this module, again if

anything biasing our overall estimates in favor of the human capital

hypotheses. Each of the continuous human capital factors (schooling,

experience, and SVP) were evaluated at -Ho around their means while

the dichotomous variables (migration and training) were evaluated at

zero and one.19

While the human capital factors were allowed to

vary in this Way, the values for the stratification, industry, and

working conditions variables were set at their respective means. TwO

ad seriatum estimates were made: one for differences in s'..hooling

alone (ED-interval) and one for the 'complete human capital module

(11C-interval). These were then compared with the earnings differ-

entials associated with the industry and stratification factors

P -

(2- interval) by computing the ratio of the 2-interval to each .of the

human capital ranges. The final numbers that result have no cardinal

meaning, but can be compared in ordinal fashion. The results are

found in Table 6.7,

The findings for white men are especially interesting. Although

the ranking of the o.cup.ation strata is imperfect because of overlapping

SVP scores, there still is a general ordinal trend in the skill

colitent of jobs as one moves from occupation group 1-3 to stratum 15-17.

0cc group 6-9 is the one major, exception to this ranking primarily

because its SVP range.is so broad (SVP=3.5-8.0). If we delete this

19To simplify the analysis the evaluation was done only for

workers whose education was received outside of the south (i.e.School' south = 0).

Z.84

267

TABLE 6.7

Z/HC RATIOS BY OCCUPATION STRATA

Oc ipationStratum

White Males

Z

IntervalED HC

Intervala Ihtervalb Z/ED Z /HC

1-3 45.87% 13.84% 13.84% 3.31 3.31

5 23.82 21.83 28.61 1.09 .83

6-9 41.49 15.25 15.25 2.72 2.72

12-14 28.42 18.24 42.82 1.56 .66

15-17 12.40 42.62 114.86 .29 .11

All Strata 21.93 37.15 97.90 .59 .22

Black Males

1-3 75.80 2.84 55.32 3.64 1.37

5 74.85 15.01 26.06 4.99 2.87

6-9 49.23 16.08 35.21 3.06 1.40

12 -14 66.83 14.74 50.58 4.53 1.32

15-17 ha , na na na n/

All Strata 61.82 23.21 83.60 2.66 .74

aThe ED-interval is the earnings range expressed in percentageterms and estimated by evaluating each regression at mean values forevery variable with the exception of education (years of schoolcompleted) which is evaluated at ±lo around its mean.

bThe HC-interval is imated ad seriatum with all non-human

capital variables evaluated at their means and the human capitalfactors evaluated at ±-10for continuous variables and zero an,s1

for those that are dichotomous.

268

TABLE 6.7 (Continued)

White Females

Occupation Z ED HC .

Stratum Interval Interval Interval Z/ED Z/HC

1-3

5

6-9

12-14

15 -J7

All M.-rata

1-3

5

6-9

12 -14 na na C na / na na

15-17 na na na na na

41.54% 15.53% 15.53%1

2.67 2.67

54.63 11.58 11.58 - 4.72 4.72

35.44 - -- --

49.63 19.264e 39.11 2.58 1.27

28.22 ' 59.71 59.71 .47 .47

58.22 1Q.17 19.04 5.72 3.06

Black Females

72.47 23.24 52.60 3.12 1.38

69.78 30.35 51.44 2.30 1.36

98.31 16.71 "16.71 5.88 5.88

All Strata

1-3

5

6-9

12-14

15-17

All Strata

110.27 13 65

64.70 13.58

84.61 10.92

40.79 2. 35

36.47 44.57

59.56 32.08

63.07 30.33 64.10 2.08 .98

All Race-Sex Groups

42.53' 8.07 2.59

34.3U 4.76 1.89

10.92 7.74 7.74

53.24 1.83 .77

110.20 .82 .33

92.88 1.86 .64

M

269

special case, we find a monotonic increase in the size of the human

capital interval as we move from the lowest skilled occupations to the

professionals category.$ In occ stratum 1-3 the total HC-interval is

a mere 13.8 percent while it reaches almost 115 in the 15-17 group.

Roughly the opposite trend is seen in the earnings differentials

associated with industry and-stratification factors (Z-iterval).

The largest Z-interval is found in the, lowest skilled - category while

the smallest is found among the professionals.' Consequently there is

a combined trend toward smaller 2/HC ratios as one moves to higher'$

occupation strata. In the lower skill groups the largest differences

in earnings are associated with differences in industry and occupational

attachment while differences in human capital begin to play a

relatively much more important role only on the higher rungs of the _

,skill hierarchy. In the lowest skilled occupations the earnings

intrval due to non-human capital factors is more than three times1

as great as the range due to schooling, skill, and experience while

amongiprofessionals the size of the Z-interval is only 1/10 that

associated with human capital. Over all strata, those workers with

la more schooling, experience and SVP as well as geographical

mobility earn almost 'double (97.9%) the annual salary of workers

in similar industries who have la less education, experience, and OJT

than average and who have never migrated since childhood. Compared

to this range, differences in industry and stratification variables

generate an earnings interval only 1/5 as large. Thus clearly for

the white male workforce as a wholt, the primary factors determining

270

the distribution of earnings are related to human capital. '

Nonetheless, for those "entrapped" in the less skilled sector,

industry:aud'stratification factors are by far the more important

. variables. As long as the entrapment continues, increases'in human

capital will have little realized value.

Among back males the results are more ambiguous. There does

not appear to be any clear-cut trend in the size of the human capital

induced wage intervals over the range of occupations nor is there/a

trend in the Z/HCratios. While black men have relatively la ger

earnings differentials associated with industry and stratification

factors, their HC-intervals are correspondingly largerrleaving

relatively smaller/Z/HC ratios than white men in occupation strata

1-3 and 6-9. On the other hand no single stratum ratio is below

unity suggesting that even in the relatively skilled strata the

non-human capital factors play a substantial role in wage determination.

The Z/HC ratio of .74 across strata discloses that both human capital

and institutional factors are each of critical importance.

For white women the non-numan capital factors ,clearly dominate

the picture with a possible exception in the professional strata.

The human capital induced intervals are universally small in the

first three occupation groups and fact in the 6-9 stratum human

capital differences have absolutely no effect on wage differentials

at all. All explained variance in these earnings are a function of

industry and occupational attachment, the Z/HC ratio being mathe-

matically undefined. Finally for all strata combined, the industry

tr,

43-

and stratification factors measured ad conjunctum are more than

, three times more powerful than the human capital variables measured

ad seriatum. Thus we move to the very opposite of the continuum from

white men, suggesting that human capital differences are relatively

insignificant in determining the female personal distribution of

earRingswhile non-human capital factors dominate the field.

The results if or black women are similar to those of black men

with the exception of occupatisn group 6-9. The human capital

intervals are of generally the same magnitude as the Z-intervals-

in each of the individual occupation zroups and across all strata.

' AgainAt appears that both hinuan capital factors on the one hand and

industry and stratification factors on the other play important roles-,

in the wage determination process. Changes in either set of factors

Can be expected to hive a substantial impact' on estimated earnings.

In concluding we can turn to the results for the whole labor

force taken together. Here we find general trends which parallel,

i,

those for white men, but levels that are much closer to those found

for-each of the minority groups. With the exception of the non-

comparable 6-9 strata, there-is a monotonic downward trend in theI

1

z-interval accompanied by a -less regular upward trend in the impact _

\6of the human capital module. Together th y produce a concise

picture of the relative impact of the two sets of factors. Among

the least skilled workers in the economy, industry and stratification

factors producean earnings differential 2.6,times the size of the

human capital interval. This ratio falls (with the obvious exception

Z89

1

272

of occ group 6-9) until it reaches .33 among the highest skilled

occupations. Again this leads us to the conclusion that human

capital factors are of substantial import but primarily only in the

higher skilled strata. For the rest of the workforde, institutionalr ij

and stratificationfactors E.ie unambiguously important as independent,

and to a great extent primary, determinants of the persdhal earnings

distribution.

For the labor force.as a whole, taking into account the relative

population size in each strata, a comparison of the Z and HC

intervals in the "grand pooled" regression suggests that the estimated

impact of the industry and stratification modules is about two-thirds

the size of the effect of the human capital module. Both are important

° with .human capital having a-slight edge.20

Nonetheless the masSive-

-earnings differentials associated with (1) industry °and occupation .

crowding (2) differences in industry characteristics and

20It should be emphdsized that the range over which the human

capital factors are allowed to vary is by no means narrow. In the"grand pooled" regression, the 93 percent earnings differential isthe' total interval between two workers who have the following humancapital characteristics

+HC-interval -HC-interval

School =13.68 years(Junior College)

Institutional Training

Iligrant

40 years Experience

SVP = 6.73(2-4 years of OJT)

School = 7.86 years(Elementary School)

ft

No Institutional Training

Non-migrant

16 years Experience

. SVP = 2.97(30 days-3 months OJT)

273

A3) miscellaneous factors concerning imperfections in the functioning

of labor markets including pure discrimination, information barriers,

and lock-in effects are obviously too large to ignore. Contemporary---

labor markets do not appear to be particularly efficient in matching

workers with given endogehous productivity characteristics to jobs,.

requiring these talents. After controlling_ for human capital as best

we can, the evidence points overwhelmingly to the fundamental soundness-

of institutionalist and stratification hypotheses and provides

substantial evidence of the superiority of the personal earnings

distribution theory presented here.

VThe implications of these findings for manpower policy and

0

particularly the row -wage workforce are far-reaching. It is to this

matter that we next turn.

CHAPTER VII

A

CONCLUSIONS AND IMPLICATIONS

This study began with.ja relatively specific concern: to under -'

stand why millions of full-time workers earn so little that their

families become "working poor" in terms of the Bureau of Labor

.

Statistic'ebudget for "low Standard of living" or worse yet the

Social.Security AdministratfOn's poverty line.1

Even more,specili-

. . . Acally_ our concern was to determine

.

towhat_extent the low incomes ot-

-the working-poor are primarily the esult ofd inadequate human capital.4 ,

a

-- vs: the legacy of labor market imperfections.

Inevitably this relatively_narrow problem gave way.to much0

0 ,;

. , .

4,,

-_'b questions about the determinants of earnings for the labor

force-as a whole -and finally. prompted the Construction of a general

distribution theory, and the deyielopment of a.comprehensive data set

to test it. Ifhile,the results of our inquiry are, of course, not

11n 1967, the Social Security Administration's "poverty line"

-for a non-farm family of four was'$3,410 while the Bureau of LaborStatistic's "low standard of living" budget for. an urban family offour was $5,915. These figures can be found in U.S. Bureau of theyCensus, Current Population Reports: Consumer Income, "Characteristicsof the Low-Income Population 1970' (Washington, D.C.: Gov't PrintingOffice), Series P-60, No. 8], November 1971 Table M., p. 19 and U.S.Department of Labor, Office of Information, "Three Standards of Livingfor an Urban Family.of'Four Persons, Spring 1967" ("rashiogton, D.C.:Gov't Printing Office), March 1969.

tiz

274

- 292

a

275

absOltitely incontrovertible, the evidence from the regression analysis

seems more than sufficient to warrant some.important conclusions

about the American labor market and particularly about the market for

less skilled labor. There area number of policy implications in

the manpower area that folloW,from this analysis. .

46

.Jt would be highly' repetitive to recap all of the results

presented in Chapters V and VI, but we can reitetate the major con-.

clusions of those chapters and comment on some of the implications "

that follow from them. .Obviously with the space available we can-only

,-,outline some of these implications. A more in-depth analysis will have

to_ wait for another day.

By far the most important conclusion of our analysis is that the

American labor market is considerably inefficient in terms of matching

What we have called "endogenous productivities" to marginal prpducts

or wages. ,,Much of the, labor force appears to be paid at rates not

consonant with their measured human capital.? The result as"relative

underemployment" of large segments of the labor force, particularly

among minorities and less=skilled workers. Without altering an

individual's human capital it is often possible, at lest hypothetically,

to increase that worker's earnings significantly by only "relocating"

2We should stress the term "measured" once again, for-it isalmost certain that some forms of human capital have not been included

in this analysis which partly account some of the unexplainedvariance in-earnings., In addition we should note that individualpreferences have not been etxplicitly taken into account so that factorslike I'voluntary" immobility may also be responsible for some of the

apparent "inefficiencies" in the labor market..

"4:93

276

the worker from one 'industry or occupation 66 another. The wage

intervals we discovered for similarly qualified workers..are large

'enolph to make the difference between poverty and a so-called adequate

family income, roeexample, when -we hold; human capital constant for

occupation group 1-3, where many of the working po6r are found, we

find a wage range of $3037 to $6386, figures that biacket the poverty ,

yline and the BLS "low standard of living" budget. In other occupation

strata we find large 'human capital constant" wage intervals as well:

$2516 ins occupation grodp 5, $3202 in group 6 -9, $2350 in group 12-14,

and $2974 in the highest skill category. For the full-time workforce

-as a whole, the wage range due to differential industry-and

occupational attachment_ii $7614 vs. $4784. Thus a worker in thp laBor

-having "average" amounts of human capital but who gains access

to a "permissive economic environment" will earn almost 60 percent

more than a similarly qualified worker in a minority-crowded,

competitiVe, unorganized,- low profit-industry.

What is also clear from the. analysis is that different segments-

of the labor force face very different problems in the labor market-

In general, low :Incomes among white men athe result of inadequate

human capital, although imperfections in labor market information and

- , , .

possibly the "lock-in" effects of prohibitively expensive geographical

.relocation and non-vested seniority and pension rights appear to

promote significant wage differences among less skilled workers. Among

_white women, on the other hand, measured differences inn human capital

can explain practically none of the large wage differentials even among

4N-

Z94

277

relatively skilled strata. Our analysis indicate's, in fact, that 95

percent of the difference in earnings between white women and white

men is due to factors other than measured human capital. Much of the

total variance in our analysis is left unexplained, but that which

can be determined is disproportionately caused by imperfections in

the job market. The segmentation of the labo r market into "male"

and "female" job slots seems to play a crucial role in wage determina-A

Lion. For blaeA men and women, both the human capital and institutional

hypotheses are borne out in the wage determination process:

In theoretically specifying "imperfections" in the labor

market, emphasis was placed on the "crowding" hypothesis. As we=

expected, the evidence for crow-ling is substantial although not

definitive. To prove_crowding as a culprit in the wage determination

process, it Wouldhave been necessary to estimates -of-,.-: .

. .

the labor- supply and demand functions in=each industry and occupation._

Unfortunately, for all practical purposes, this is an impossible task.

The minority, employment variables we chose as proxies have the problem-.

of being substantially colinear with race and sex particularly because-r

sex-linked stratification is so pervasive. Never'heless the "crowding"

_

variables were often significant within individual race-sex groups

(including white men) after controlling for human capital and even

in a number of the cross race-,sex equations after dummies for race and

sex were added. Both of these tests suggest that crowding has an

independent effect on earnings.3

What is important to remember,

3Data from the Census Bureau's Consumer Income series provides

278

however, is that whether the STRAT mOdUle measures the specific forh

of discrimination known as "crowding" or some other form of discrimi-

nation is less important than the fact,that something to do with race

and sex is an extremely powerful determinant of personal income. ItO

is perhaps the,major reason for the lack of colinearity between

endogenous productivity and earnings.

'`Insofar'as there is evidence of 'crowding" it was possible to

divide its effect into "simple" and "complex" forms depending on

whether in addition to the stratification factors differences in

industry choracteristica had an impact on the distribution of earnings.

;The - compelling conclusion' seems to be that "complex" crowding_is the

general rule throughout but particularly so for the minority groups. I

Simple crowding explained wage differentials"for white men in occupa-// _

tiOn Strata 1=3 and 6-9 while-all black male, black female, and white

female groups (excluding-white women in occ' group 1-3) were typified

by significant industry as well as stratification variables.

Employment in a permissive economic environment of extensive oligopoly,4

high profits, and capital intensity added significantly to the earnings

corroborati'Ve evidence of "crowding" of white female labor. Since1955 'the ratio of full-time, full-year white female/white male wage andsalary income has secularly fallen as the labor force participationrate of white women has risen. In 1955 the ratio was .644; by 1968 ithad fallen td .586 and is continuing-to fall. In the absence ofcrowding--and assuring no divergence in human capital or intensificationof pure "sexist" attitudesithere is no reason to believe this ratiowould fall. The increase in female supply should affect white malewages as -well, if crowding is not operating. No other theory seems to

explain this phenomenon as well. See U.S. Census Bureau, Current .

Population Reports -- Consumer Income Series, P-60, No. 69, April 6, 1970,Table A-8, p. 86.

96

t

279

of minority group members, a finding in complete accord, with

Aiitibn 1-institutionalist theory.

0"e thing that becomes abundantly clear in the analysis, parti-

cularly in Chapter VI, is the dramatic change in the relative

importance of the human capital and non-human capital factors as one

moves from the low-skilled to the high-skilled occupation strata;

InduStry and stratification factors are universally dominant amo'g the

lower-skilled strata while human capital takes on a larger and larger

role as one proceeds up the occupational hierarchy. For the labor

force in occupation group 1-3 we found that the industry and stratifi-,

_cat-ion factors produce an earnings differential 2.6 times the size of

=

the-wage differential due to differences-in human capital. But among

the-highest skilled group tbe ratio falls to only .33 after a near

secular decline through the whole occupational range. At the top of

the hierarchy labor markets appear much more "efficient" in

allocating workers accordihg to their endogenous productivity

'characteristics.

There area number of more specific findings that bear repeating.

0-e of these is the statistical insignificance of institutional

training as a determinant of earnings for every group with the

exception of black males. For black men,

significant and relatively substantial in

12-14 in addition to the cross-occupation

the training variable was

occupation 'groups 1-3 and

regression. In no other

regressionwas this true. This may be due to the poor measurement of

this variable, or it might have some important content as we shall

144'*7' j

a.,

280.

later suggest.

On-the-job training, as measured by SVP, is an especially

powerful variable in the higher occupation groups and across all

-occupation strata, but there remains great confusiOn as to what this

finding actually proves. Because SVP is only obtained after access,

to a speclfic occupation is gained, leis difficult to treat it in

like manner to the other variables in the human capital module. If

occupational access is barred by discrimination or some other1

imperfection in the labor market, SVP may be better treated as a

stratification variable and its eff

1

ct counted here. On-the-job

training is therefore of critical portance in wage determination,

but it is difficult tosuggest h.Ow ocial policy might be develbped_to

deal win- it based on our analysis.

Finally we should note that we have found practically no evidence'

of "compensatory" wage payments for physically demanding, unpleasant,

or_dangerous work. While otir proxies for these factors are not

especially well- measured, we often find a negative rather than positives

sign on these variables. If we follow the signs on the coefficients, 0

, of.

we note that the-Fe area number of negative signs in the low-skill

, categories followed by inJignificant coefficients in the middle and

higher skill categories. The, one positive sign we find is among

relatively skilled white men. Only here is the compensatory theory

borne out by the evidence. In the lower occupation strata differences_

in working conditions may be completely overshadowed by the effect of

stratification while in the upper strata the true effect can be

Z58

281

0

measured because stratification plays a much' weaker role.

Alternatively it is possible that'differences in ability and skill

have not been held constant enough to pick up small, but nonetheless

existing, "compensatory" effects.

Theories to Explain These Results

The overall picture then is one of a highly:imperfect labor

market stratified by race mid sOc.. By no means is the human capital

theory disproved or completely rejected, but the general theory of

\-

personal earnings developed` here is clearly: superior in its ability. -.

to describe the parameters of the \earnings distribution.' Yet the, T .

"theory" is primarily only .a description even ailowing'for the4, r

analytic properties of the crowdin, g,\ hypothesis, The unansvieredio .

question is what dynamic is responsible for promoting such a labor

m doA ket structure and'then what cabe'done to alleviate its perverse

distributional and allocational effects., We cannot hope to give a

definitive answer to this gargantuan question, but we can attempt,some

brief conjecture.

We should note once again for the record that a strict human

capital theorist will probably deny much of the evidence presented

here and therefore possibly not see the need for an explanation at all.

Arguing that the human capital module is.misepecified and the data

inadequate is one possible way-to explair away=the results found in\

this analysis. There is no "measure'of innate talent and admittedly

there is an inadequate specification of interactions between human

;99

4

282

capital variables. But this, in our opinion,-cannot-ccouht for. 4

the apparent wage intervals we have-found associated With non-human

capital factors, particlarly after loading the analysis in favor of

the human capital explanation at every turn. We feel that our analysis4

clearly does ddmonstrate the existence of widespread imperfections1

which cannot, be explained away so simply. Assuming our resul_s}

generally carect we need to explain-them.

One possible explanation comes from radical stratification theory.

The large wage differentials we have found associatdd with race and

sex can be interpreted as consistent with the "diyide and conquer"

theory which is currently being developed.4

At considerable risk of

oversimplifying and thereby vulgarizing radical theory, the argument

can be-paraphrased. In order to keep the whole working class from

organizing en masse to overturn the capitalist order, the "ruling

class" his consciously devised institutions to prevent the developmenta

of subjective class consciousness among all workers. Racism and

sexism have been deliberately instigated to affect divisionsNwithin

the working class along these lines. In its "vulga treatment,

radical stratification theory looks to conscious-racist and sexist

hiring and promotion decisions by management as the major tools of the

"divide and conquer" strategy. More realistically, however, radical

4,For one version of the "divide and conquer" theory see David M.

Oqrdon, Richard C. Edwards, and Michael Reich, "Labor Markr7tSegmentation in American Capitalism," Conference on Labor MarketSegmentation, Harvard University, March 16-17, 1973 (mimeo); Also see,Stephan Marglin, "What-Do Bosses Do?" Review of Radical PoliticalEconomics, Vol. 6, No. 2, Summer 1974.

300

a

283

theory points to the roles of social and cultural institutions,

-particularly the schools and "bourgeois" family customs, in dividing

4the working class.'

Taken in, this broader context, radical stratification theory, ,

welbelieve, has much 'to offer in producing An understanding of the

overall income and wealth dir we experience in the United

States It is clear that massive differences in schooling and in sex .

roles are fostered in &Lir society which end up segmenting the labor

force into different occupakdon strata.5

Race; sex, and social class,

as we argued in the general stratification theory, can easily be seen

as the primary exogenous factors in determining the final distribution

of income.

But the problem in the present'analysis is much narrower in at

least one respect. Here we have held human capital constant and

asked they question how much of the variance in earnings can be

,explained by other factors. We therefore need a much more specific

theory which relates these human capital constant wage differences to

-factors that operate in specific labor markets, not necessarily the

social milieu more broadly defined. One obvious answer to explain

wage differences is pure 'discrimination on the part of firms. Another

(

5For an excellent treatment of this subject,'see SaMuel Bowles,

"Unequal Education and the Reproduction of the Social Division ofLabor," Review of Radical Political Economics, Vol. 3, No. 4, Fall-Wintet 1971 and Samuel -Bowles, "Understanding.lhequal Econoric

Opportunity: The Role of Schooling, and Family Economic Status,"American Economic Review, May 1973.

4).

301

284

is simply infopmat;on Imperfections which never get fully resolved.

But there is good reason to believe that such answers are

indeed too simplistic. An adequate theory.must do more than expltin

the existence of wage differentials which are not related to human

capital. Such a theory must also be able to explain why industry-

related wage intervals are largest in the lower-skilled strera while

at the top of the occupational hierarchy the huMan capital elements-

dominate. Extending the brief analysis in Chapter VI, an eclecticI (

theory can be-suggested which meets these requirements. leis based

on a combination of theories including (1) job competition (2) labor

market search (3) quasi-fixed factor and (4) statistical discriminatione

all of which are placed within a specific historical context. A

rigorous treatment_of this eclectic model can most likely,be

demonstrated, but for the present we must be content with simply laying

out the basic structure of the argument. One thing that is especially

significant about the eclectic theory is that while it is consistent

4

with the "crowding" hypothesis and radical stratification theory,1%

it does not rely on a "conspiracy" theory of capitalist institutions.

As in Chapter VI assume a job competition model where job/wage

slots arP.given exogenously, at :least in the short'run, and there are

fixed costs of hiring and training labor. The fixed costserise

with increasing job complexity so t there is a general positive

relationship between the degree of 'fixity" and occupation strata.

Alsb assume that information about potential employees is imperfect

;.,,,and involves procurement costs. Information on average group

0

285

characteristics, even if_impreciSe, is relatively inexpensive to

obtain while information about specific individuals is costly.

With these assumptions and the'additional one that firms

attempt to minimize total labor cost in an attempt' to maximize profit,

we can generate a theory fully consistent with most of our findings.

Firms will attempt to minimize the sum of direct payrhents to labor

plus search costs plus hiring and training costs: These costs are

not independent of each other for higher wage offers can reduce search

costs by increasing the supply of labor to the firm and greater search

effort can reduce training costs by providing a higher expected

probability of acquiring workers who can be quickly and efficiently

---trained. Wherever the training requirements for a specific job-are

-minimal we can expect that the rationdl firm will find it unnecessary

to invest heaiiily in search, for recruitment "mistakes" do not force

the firm to incur,lerge sunk costs. On the other hand, wherever

.training'requirements are substantial, the Cost of a recruitment error

is considerable. Therefore we can eXpect that there will be a

positive relationship between the degree of fixity in a particular

occupation stratum and search costs. To reduce the risk of large

.unprofitable sunk costs, firms will search intensively for recruits

destined for skilled positions while expending little search effort for

workers who are hired primarily to fill unskilled (low "fixity") slots.

In more specific terms, firms will investigate the individual

characteristics of their prospective skilled employees while using

rules of thumb or general search strategies to fill unskilled job

303

_

I

286

slots. In the latter case, many firms will resort to statistical

selection of one sort or another using supposedly objectively

perceived group characteristics in making hiring decisions about

specific individuals.6

While "rational" in the limited economic'

sense, such a strategy is ob'lliously prejudicial by definition.

One important question, pf course, is what group characteristics

are used for screening. Here is where the historical context

inevitably plays a'crucial tole.- The social and cultural institutions

--and belief syStems embedded in any society are marked by,substantial

inertia. 'Onqt firmly established for whatever reason, they tend to

be passively, if not actively, perpetuated. Without reviewing American

(or for that matter much of all Western European) history it seemsI

hardly necessary to "prove" that both blacks and-women have

hlstorically been relegated to disadvantaged positions in the labor,

force, blacks through involuntary servitude and racial segregation

and wren through family custom.7

Through the years custom and habit0

1.have produced some objective diffeiences in group characteristics as

well as (and, probably more importantly) induced lingering perceptions

of differenc4s which may have no basis in fact. Both 9f these no doubt

have a s bstantial impact-on recruitment patterns.

a \I

6See Kenneth Arrow, Some Models of Racial Discrimination,') op. cit.

7This analysis obviously begs the real question: why did he

racial and sexual institutions and beliefs develop in the first, place.Here "divide and conquer" theory suggests one possibility.

304

287

In the context of our culture, statistical screening then works4

itself out in terms of racist and sexist hiring procedures, not

necessarily out of an express desire to "divide and ednquer" or out of

a deep-seated commitment to white male domination of society (although,

both of these may be operating). Rather if firms have widespread

beliefs'about the expected probabilities of employee "success"--

whether these expectations be groUnded in fact or not--the result will

be stratification of the labor force.

t

To review, the lower the degree of fixity, the smaller the

potential cost of a recruitment mistake.Which in turn leads to

minimal search effort and a general tendency toward statistical

-disCrimination as the firm's search technique. The end result

inevitably is stratification of the labor force in the lower occupation

.47

strata. If a sufficient number of firms screen on the same character-.

istics, the result will be crowding and the development of large wage

differentials between groups in the economy. Once the initial

stratification has taken place, differential supply of on-the-job

training (SVP) may tend to exacerbate these differences. Also once

this system has been generated, it tends to be perpetuated. If the

screening procedufes seem to have "worked" in the past, they will tend

to become rules of thumb to follow in the future. Thus even without

pursuing a Conscious policy of "divide and conquer," a private cost-

minimizing system will tend to perpetuate non-human capital linked

stratification as long as labor market informati,on is imperfect and '

costly to secure. In the absence of government intervention, the

305

288

"social costs" of stratification will continue to be borne by

minority membersof the labor force. And these costs, asme have

amply shown, are often immense. Whatever private gain mii;hi come

from a "divide and conquer" policy if perpetrated may redound to the

"capitalists*" benefit even-without their active participation. In

effect, then,- a market system operating in an environment of

(1) ubstantial quasi-fixed costs for skilled labor (2) non-zero cost

information, and (3) a legacy of racist and sexist custom will tend to

produce a,"meritocracy" at the top of the occupational hierarchy and

racial and sexual stratification at the bottom.

What's To Be Done?

The labor market we_have uncovered is one involving-large scale

inefficiencies if one defines efficiency by a cOlinear mapping f

endogenous productivity characteristics and earnings. Yet we have

also posited a 'theory that the "inefficiencies" may be due to the labor

market operating the best it can given the context, of a supposedly free__"

market and limited information. If the market is to be moved toward

a more socially "efficient" and equitable allocation of labor, what

must be done?

Obviously a labor market which is segmented in such a complex

manner as we have discovered requires a multifaceted set of policies

to ensure equal opportunity in the labor force and reduce allocational

inefficiency. Wo single policy will be sufficient to redress the

stratification in the labor market. Without going into detail, we can

306

289

lay out a few areas in whith we feel policy must be directed.

One obvious finding is that although labor market stratification

is widespread, 'differences in human capital are still extremely

important. Within each race-sex group increased human capital in the

form of 'formal education, labor force experience, on-the-job training,

and migration all pay off interms of higher earnings. Yeethere are

great disparities that, remairi in the allodation of human capital \

between individuals particularly on the basis of race and class. This

8

QO

To redress the balance requires at a minimum equal. educational

has been amply demonstrated by other researchers.

opportunity if not compensatory educational programs for groups which

have historically been at a competitive disadvantage. In order to1

ensure equal labor market opportunity may in fact require unequal-4

educational opportunity, discriminating in favor of previously

discriminated against minorities. Quota systems and direct application

of affirmative action in college admissions, for example, are probably'

required. Other forms ofOluman capital may be equalized by providing

relocation allowances for those who can profit by moving from one area

8See for instance, Samuel Bowles, "Schc:ioling and Inequality fromGeneration to Generation," Journal of Political Economy, May 1972; W.Lee Hansen and Burton Weisbrod, "The Distribution of Costs and DirectBenefits of Public Higher Education: The Case f California, Journalof Human Resources, Spring 1969; Stanley H. Ma ters, "The Effect ofFamily Income on Children's Education: Some Findings on Inequality, ofOpportunity," Journal of Human Resources, Spring 1969; James S. Coleman,et al., Equality of Educational Opportunity (Washington: U.S. GovernmentPrinting Office; 1966); Patricia Cayo Sexton, Education and Income (NewYork: Viking Press, 1964); ancb Samuel Bowles, "Towards Equality ofEducational Opportunity?" Harvard Educational Review, Vol. 38, Winter

1968.

O

3 o7

P

290

to another and providing incentives for firms to ,give on-the-job

training to minority group'members.9

But the primary implication of bur analysis is that manipulation.

of non-human capital factors 16 also critical to addressing current

labor market Problems, particularly of the working poor. Insofar as

direct discriminatiOn is still widespread in the labor market, it is

clear that equal employment opportunity legislation must be extended

and forcefully implemented. As in education; affirmative action in

employment is an important tool in promoting social efficiency in

the labor market: More recent implementation of affirmative action

may have already begunyto excise the wage_differ,pntials betc.feen race-

sex groups. Clearly_such a direct approach to ending discrimination'

is warranted by the,results presented in our-analysis.

Beyond direct affirmative action, there seem-to be a number of -

roles the government can play in regard to statistical discriminatiori.

If,_as we suspect, screening is ,oftgn baSed on erroneous conjecture

"about group characteTistiCaplyegovernment can help to "correct the;

\-record." Such intervention in the market would not have the same

g

powerful effect f direct action, but it no doUbt should be in the-

ernment's policy tdol-box. Where substantial,discriMination is

ective," then the government must find the means for decreasing .

the rivate sector's cost of procuring individual job applicant

I

This latter policy suggestionshould be qualified foraccordingto the General Accounting Olfice, firms can take advantage of on-the-job training subsidies without providing much additional benefit todisadvantaged workers. See Chapter I, foptnote 8. . °,

3G8

291

information. This role could be played by a much more effective

public employment service which would have the funds and the

expertise to accurately screen individuals on the basis of relevant

job characteristics.10

The current Employment Service-has generally

(

failed in its attempt to bring low-skill workers and jobs together.11

More specific implications can be drawn from the analysis about

manpower training programs. We noted at the beginning of this inquiry

the: most social scientists and government officials have been

disappointed with the performance of manpower programs in the United

States. In the present analysis we find additional evidence that

institutional vocational programs have failed td have much of an

impact on earnings (althoUgh we-have.no information on their effgt--

on securing employment). The_ "training" variable is_significant_only__

for black males as far as the individual-race-sex equations are con-

cerned. Training is never significant for white men nor eithef group

of women. This result caused some consternation for we originally

suspected that if any group should benefit from institutional training

10See Richard Lester, Manpower Planning

(Princeton: Princeton University Press, 1966)Manpower and the Public Employment Service inYork State Department of Labor, 1966).

in a Free Societyand Alfred,L. Green,Europe (New York: New

110ne new picoe of evidence for -this conclusion comes from a

recent'study in the. Boston labor market conducted by the Social Welfare:Regional Research institute, Boston College. In virtually p11 of thefirms studied by SWRRI, the employment service was not considered areliable source for obtaining relatively less skilleliqabor and there-fore was rarely contacted about job vacancies.' Robert Hubbell-rand

Martha MacDonald, "A Study of Employee 'Recruitment _in Boston," SocialWelfare Regional Research Institute; Boston College, Workirig Paper,forthcoming.

309

vi

292

it would be white men. But here we find only black men apparently

gaining from these programs. One explanation, of course, is that the

result is purely spuriou, but there is an alternative that we prefer.

It has often been suggested thatomanpower training programs do

'little to increase the actual productivity of workers but play a

primary role,in the screening of recruits.12

White males do not gain

from this additional "screen" because they already are "screened into"'

the better occupations and industries by reason of their race and sex.

For blacks,_ however, training-plays the critical role of signaling to, 41

potential employers the special motivation that trainees may engender

. or appear to engender. In this case firms can use enrollment in e

training program as a way of screening in a few black recruits whhef

1.,-,-_-the normally operating racially-linked statistical discrimination ,_:-,.,..0

.1

cz: __- . ,

-sort-ens out all others. If this_is true, then institutional training

is obviouslyen important "human capital" variable for black men,

although its usefulness as a screening device might depreciate as thett

number of institutionally trained black workers increases. All of this

is but conj- ecture at this point, but it makes some sense within the

13context of the general stratification theory underlying our analysis,

12See Ivar Berg, Education and Jobs: The Great Training_Robbery.

(New York: Praeger, 1969).

13It has been suggested to me by several colleagues that *insti-

tutional training may even be a negative credential for white men ifemployers see enrollment -as an indication of labor market disadvantage-

mentA A "'good" worker should not,need a vocational training prOgram,might be the thinking of employers.

310

293

Finally we may conclude by mentioning the broadest implications

of-our analysis. What we believe we ultimately have shown is that

the distribution of earnings in the United States is substantially

arbitrary with respect to human capital. A large part of earnings

differentials have been shown to be related to non-human capital

factors so that the overall distribution of earnings can be described

as "unfair" with equal human inputs being rewarded with vastly unequal

returns. Much of the- justification for existing wage and income

differences ittributed to marginal productivity theory thus pales

before this analysis even if one has accepted the questionable premise

that a just distribution of income is one based on marginal products.

Imperfections are so extensive and their effect so deep that the

14

relationship between endogenous productivity and marginal product is

far from colinear.

'If individual policies of systematically countering imperfections

in the labor market cannot assure a solution to the distribution

problemwhich is very possibly the case--then it will probably be

required, at least in the short run; to resort to direct redistribution

- -of-income- via- negative income taxes or other forms of-income guarantee.

Such a redistributiOn would be far from Oerfect in-redreaSing "the

balance, but would be in general accord with the--policy implicationS

that flow from the crowding hypothesis. Under a negative income tax,

14See J.B. Clark for an early statement of the neoclassical

"just" wage doctrine. John Bates Clark, The Distribution of Wealth(New York: MacMillan, 1900), esp. Chapter 1 or 'Milton Friedman, PriceTheory: A Provisional Text (Chicago, Aldine, 1967), Ch. 10.

311

O

294

income would be.transferred from those who have higher earnings

in part due to segregation to those who have.been the victims of a

stratified labor market. In this case direct redistribution is a

surrogate for what would actually occur in the labor market if

barriers to mobility were reduced.

Thus the policy implications of our findings are extremety'far-,

reaching. They demand that policy-makers understand the need for-

wide7ranging intervention in the economy at the micro level in order

to move toward a more "efficient" and distributionally fair labor

market. Direct attacks on the structure of labor markets will often

_be much more effective particularly for lower-skilled workers than

attempts to remedy all problems through individualistic human capital

policy. All of this, of course, abstracts from even broader questions

of the control of the economy at large . . . but this is a question

to which I hope to devote my future work.

312

APPENDIX -A

THE DATA SOURCE

O

APPENDIX A

THE DATA SOURCE

The ability to test the general earnings model developed in

this thesis rests on the development of a comprehensive micro-macro

merged data set. This data set which is based on the 1967' Survey of

'Economic Opporthniq. was prepared over the course of one and a half

years by Professor Mary Stevenson, now of-the University of

Massachusetts-Bos'ton, and myself with the help of the staff at the

Institute-of Labor and Industrial Relations-Research Division at the

University of Michigan.

Three types of data were compiled, and merged to complete the

1 computer file, S1480,'which was used in all of the analysis. Data

from the-1967 SEO formed the basic sample. Information on personal

-characteristics flom this source wa,g suplacmented by-linked information

-from the panel portion of the 1966 SEO. The second type_of data

involved the compilation and linkage to. the SEO file of data on

=

occupational characteristics, the primary sources being the Dictionary

of Occupational Titles and the U.S. Census volume on "Occupational

Characteristics." The third type of data was macro information on

industry characteristics compiled from a broad range of sources. This

too was merged onto the SEO file so that the final 51480 tape contained

personal information on a 'large sample of workers with accompanying

296

314

297

0

data for each individual on that Worker's occupation and industry.

_

Thuis for every worker on the SEO file'we can ascertain such individual

characteristics As educational attainment and age, but also a-,goodO

deal about the training required.and the working conditions in that

worker's occupation and the profit rate' and the concentration ratio

--in that worker's industry.

The SEO Data Base

The Survey, of Economic Opportunity (SEO) for\1967 contains

information collected in a sample survey conducted in the Spring of

that year.1

The Bureau of Census conducted the survey for the Office

of Economic Opportunity to supplement the information regularly

collected in the-Current PopUlation Surveys (CPS) for February and

March of each year. The common items in the SEO and CPS include

personal characteristics (e.g. age, race, sex, family relationship,

marital status), last year's work experience and income. The CPS,

conducted every month, is the basis for monthly labor force statistics

including unemployment rates.

In addition to the normal information gathered in the CPS, the

SEO provides information on dimellsions of poverty not usually obtained

between the decennial census years. Of equal importance, the SEO

oversampled nonwhite census tracts to achieve better estimates of the

Ihis file is available from the Data and Computation Center,The University of Wisconsin, 4452' Social Science Building, Madison,Wisconsin 53706.

315

298

characteristics of the nonwhite poor. The national self-weighting

sample has approximately 18,000.households, drawn in the same way

as the CPS sample. The oversample in areas with,large nonwhite

populations has 12,000 additional households. The combined sample

yields over 57,000 adults, age 16 and over.

Working 40 weeks in the previous year and a minimum of 30 hours

in the survey week at one's "usual" job. qualified an individual as a

full-time full-year worker for the purpose of the present analysis.

Individuals who worked at least this much, but responded that their

present job was not their "usual" job were not included in the sample,-

for the hourly wage they report may have little relation to the

occupation and industry indicated in their survey record. All whites

and blacks in the self-weighting sample 25 years or older who

,qoalified as full-time full-y,c5i workers by these criteria were _

Automatically included in our analysis. Nonwhites who are not black

were excluded from the sample, since including Puerto Ricans and

Mexican-Americans in with the white or black samples would be

misleading because of different labor force characteristics.

Analyzing them separately, however, is impossible because of small

sample size.

In addition to the self-weighting sample, blacks in the supple-

mentary survey who met the full-time full-year qualifications were

included in the study population. The blacks in the supplementary

sample were found to be quite similar to the blacks in the self-

weighting sample according to an examination of means and Chi-square

316

299

tests. Thus, combining the. two samples did not significantly

.alter the qualitative results and the additional size added

immeasurably to the reliability of the estimates of black character-

isEics. However, the whites inithe "nonwhite" supplement were not

included in the analysis. Preliminary investigation indicated signi-

ficant differences between whites in the self-weighted survey and the

supplementary sample. Without the use of weighted counts, the addition

of the whites in the supplement would have qualitatively altered the

results for all whites in the population. Without the supplement,

the sample size for the white population is sufficient to yield

reliable statistics in even highly stratified analysis.

In all, there are over 57,000 individuals 16 years and older in

the 1967 SEO file. Of these, nearly 14,000 meet the full-time full-,

year and "usual" job qualifications. (See Table A.1) -Almost 10,000

are in the:self-weighting sample; the rest are blacks in the supple-

mental survey. Nearly 9,300 are males; 4,700 are females. In all,.0

there are nearly 9,000 whites and 5,000 full-time full-year blacks.

White males are the most numerous; black females are ldast numerous.

Nonetheless there are 1,867 full-time full-year black women workers

in this study. In all four race-sex groups it is then possible to

stratify the sample extensively and still maintain ample cell sips

and reliable estimates even after discarding individuals for whom

there is incomplete occupation and industry. data.

The 1967 SE0 yields information on all of the individually

measured variables needed in the analysis with the exception of a

313

300

TABLE A.1

THE SE0 SAMPLE POPULATION OFFULL-TIME FULL-YEAR WORKERS

Self-weightingSample n

NonwhiteSupplement Total

White Males 6257 6 6257Black Males 594 2242 3036

Total Males 6851 2442 9293

White Females 2736 0 2736Black Females 362 1505 1867

Total Females 3098 1505 , 4603

.

Total White 8993 0 8993Total Black 956 3947 4903

Grand Total 9949 3947 13896

measure of vocational training. A question about institutional

training was asked, however, in the 1966 SEO. Given that he same

addresses were interviewed in both years, a large percentage dl the

sample population wen- interviewed in both years. Given this fact

and the ability to link the 1966 and 1967 SE0 surveys, it was possible

to add information about vocational training to over three-fourths

of the 1967 sample. This was done.

The final variables used from the Survey of Economic Opportunity,

included:

Hourly Wage - The dependent variable is hourly wage (last week)

computed from answers to questions which indicate the number of hours

respondent worked last week and indicating the respondent's earnings

311

301

last week. Given the proviso of full-time employment at "usual"

job, this variable is somewhat cleaner than comparable measures of0 4

'earnings. The major problem is the possibility Of inflating the

straight-time earnings rate due to special overtime rates. To the

extent that overtime is "usual" in a given occupation, this would not

necessarily overstate the earnings measure we are after. The major

problem arises when in the week before the interview, an individual-

worker had an unusual number of employed hours. It was not possible

to take this factor into account. However it seems reasonable to

believe that it would not seriously bias the final results.

Schooling - This variable is taken directly from the question on

"highest grade completed.

School-South - This interaction term is computed from the answer

to the question on highest grade completed and the answer to the

-- question "region at age 16." A dummy value of 1 is assigned to the

region South.

-Trainia& - This variable from the 1966 SEO,was intended to

.account for institutional training of a variety of types. These

included: business college or technical training, alrenticeship

training, full-time company training, vocational trai n the armed

forces, other.formal vocational or technical training, of non-regular

schooling. A dummy value of 1 is assigned'if the.respondent had been

or was currently enrolled in any of these training programs. It was

not possible to ascertain whether the respondent had completed

training nor whether the training was related to subsequent occupational

.319

302

attachment.

Migration - This dummy variable was computed from the question

"age left last residence." A dummy value of 1 is assigned to a

respondent who was a "nonmover" indicating that this individual had

never moved more than 50 miles from his residence since age 16.

Experience - This variable was computed from questions on the

individual's age and highest grade completed. The actual variable

used was {AGE- SCHOOL -6) indicating the number of years out of school

and presumably available to participate in the labor force.

Union Member - This stratification variable was taken from the

question, "belong to a union." It was asked_only of private wage and'

salary workers. A union was construed as an organ_ted craft,

industrial, or professional union such as the bricklayers union, the

United Auto Workers, teachers union, etc. Fraternal or.social

organizations such as,Masons, Shriners, Elks, etc. were not considered

as unions. Similarly, professional organizations were not considered

as unions.

Race - This variable referred to "white" or "negro."

Sex - This variable referred to "male" or "female."

In addition to these factor's, the SEO provided occupation and`

industry codes for each individual which were used to merge occupation

and industry data onto each micro-record.

X320

303

Construction of Occupation Scores

Three kinds of occupation data were collected: (1) data on

education and training requirements (2) data on worPing conditions,

and (3) data on the1degree of minority employment. The first two

types were compiled with the use of the Dictionary of Occupational

Titles (DOT) while the third made use of the "Occupational

Characteristics" volume of the 1960 Population Census.2

The DOT is a vast store of data on individual occupations which

is rapidly coming into use by social scientists after being primarily

the domain of job placement officers. The Dictionary has information

on the general education and specific vocational preparation requirements

as well as data on working conditions for some 4,000 specific

occupation titles. The data for the Dictionary were collected and

developed according to job analysis techniques established by the

U.S. Employment Service. In most cases, the same job was analyzed in

two different establishments in one State and then in two different

establishments in another State. The findings of these studies were

correlated and job definitions prepared. As a result, information

presented in the Dictionary reflects the findings of-the U.S.

Employment Service from approximately 75,000 studies of individual

job situations.3

2U.S. Department of Labor,, Manpower Administration, A Supplement

to the Dictionary of Occupational Titles, 3rd. Ed. (Washington, D.C.Government Printing Office, 1966); and U.S. Census Bureau, UnitedStates Census of Population: 1960,20ccupational Characteristics,"Series PC(2) 7A, 1966.

ASuleiEpnenttothelarofOccuationales, op. cit.,p. vii.

321

304

Foftunately for economic research, the General Educational-, ,

Development AGED) and Specific ocational Preparation. (SVP) scales

for each occupation are based om functional or performance require-

inents. According to Sidney Fine,4

These are the requiremets determined by objective jobanalysis as necessary and sufficient to achieve averageperformance in the specific tasks of the jobs. Suchestimates try to focus on the tasks performed in relationto-the things, the data, or the people involved in those_tasks. ti

Thus for.each occupation an objective analysis of the job, not the

worker, is performed, ,yielding an independent measure of job content.

This is,'of course, of crucial importance for otherwise there would

be in the dataa tautoloditgi relationship between aoworker's

characteristics and the characteristics of the job he or she performed.

A "yeIrs of schooling'--t-Cale for GED was avoided when the

estimates were being developed. Instead the approach used was-to ask:

What basic-skills are pdople supposed to acquire as a result of(

general education, and what is the role of these,skills in jobs?

Three fundamental skills were delineated; reasoning,'mathematics,

and language. The requirements for each of these fundamental skills

were explored in a variety of jobs, and the scales currently in use

were developed to permit estimates from very low to very high require-.

ments. Again, attempts to translate these estimates into a time

4Sidney A. Fine, "The Use of the 'Dictionary of Occupational

Titled' as a Source of Estimates of Educational atil TrainingRequirements," Journal of Human Resources, Winter 1968, p. 365.

32249

t's

305

scale related to years of schooling completed in one educatidnal

system were resolutely avoided.5-

Each specific occupation was assigned a GED score on the basis

of these requirements. 'The rating system is reproduced in Table

reprinted from the Dictionary.

The SVP scale, unlike GED, is a time scale. It is described as

follows:6

Vocational Preparation: This is the amount of time requiredto learn the techniques, acquire informatiori, and developthe facility needed for average performance in a specificjob-worker situation. This training,may be acquired in aschool, work, or a vocational environment. It does notinclude orientation training required of even every fullyqualified worker to become accustomed to the special con-ditions of any new 4nb. Specific vocational training includestraining given in a-, of the following circumstances:

(a) Vocational education(b) Apprentice training(c) In-plant training(d) On-the-job training(e) Essential experience in other jobs

Each of these various types of training for an occupation was con-

verted'to a time Period and the periods were then adde' to provide

the,final estimate of required specific vocational preparation (SVP)

for each job. The SVP scale is reproduced in Table A.3.

In 'addition to the GED and SVP scales, the POT rates occupations

in terms of "Physical Demands" and "Working Conditions." The Physical

Demands scale includes data on required strength as well as the need

5Ibid., pp. 366-367.

6Ibid., p. 368.

323

306

TABLE A.2

'GENERAL EDUCATIONAL DEVELOPMENT

Level Rusoning Dote loptetnt Itathomotical Development Langugo Development-

6

4

$

1

Apply principles of logical orscientific thinking to a widerange of intellectual andpractical problems. 'Dealwith nonverbal symbolism(formulas, scientific equa-tions graphs, musical notes,etc.) In its most difficultphases. Deal with a varietyof abstract and concretevariables. Apprenend themoat abstruse classes ofconcepts.

Apply principles of logical orscientific tninking to defineproblems, collect data,establish facts, and drawvalid conclusions. Inter-pret an extentive variety oftechnical instructions, inbooks, manuals, and mathe-matical or diagrammatic+form. Deal with severalabstract and concretevariables.

Apply principles of rationalsystems' to solve practicalproblems and deal with avariety of concrete variablesIn situations where onlylimited standardization 'exists. Interpret a varietyof Instructions furnished inwritten, oral, diagrammatic,or schedule form.

Apply common sense under-standing to carry out in-stniction4 furnished inwritten, oral, or diagram.matic form. Deal withproblems involving severalconcrete variables In orfrom standardized situa-tions.

Apply common sense under-standing to carry'eut de-tailed but uninvolvedwritten or oral instructions.Deal with problems in-volving a few concretevariables In or fromstandardized situations.

Apply common sense under-standing to carry outsimple one- or two-stepinstructions. Deal withstandardized situationswith occasional or novariables In or from thesesituations encountered onthe job.

Apply knowledge ofadvanced mathe-matical and Binds-601 techniquessuch as differentialand integral cal-culus, factoranalysis, and

'probability deter-mination, or workwith a wide vari-ety of theoreticalmathematical con-cepts and make,original applica-tions of mathemat,Ical procedures,as in empiricaland differentialequations.

Perform ordinaryarithmetic, alge-braic, and geo-metric proceduresin standard, prac-tical applications.

Make arithmeticcalculations in-volving fractions,decimals, andpercentages.

Use arithmetic toadd, subtract,multiply, anddivide wholenumbers.

Perform simple ad.:dition and sub-traction, readingand copying offigures, or count-ing and recording.

Comprehension and expression ofa level to

Report, write, or edit artic:eefor such publications as'newa-papers, magazines, and technicalor scientific journals. iPrepareand draw up deeds, leases, wills,mortgages, and contracts.

Prepare and deliver lectures onpolities, economies, education,or science. /Interview, counsel, Or advisesuch people as students, clients,or patients, in sue matters aswelfare eligibility, Vocational

_ rehabilitation, me tal hygiene,or marital relatiods.Evaluate engineering teebnlcaldata to design buildings andbridges. ' t, -

Comprehension -vend: expression ofa level tO-

-Transcribe-dietation, makeappointmenti-for executive andhandle his personal mail, inter-view and screen people wishingto Speak to him, and write rou-tine correspondence on owninitiative. jr

:-Interviewfjob applicant/1A°determine work best suited fortheir abilities and experience,and contact employers to interestthem in services of agency.

Interpret technical manuals aswell a$/drawings and specifica-tions, such as layouts, blue-prints, and schematics.

Comprehension and expression oflevel to-

-File, post, and mail such mate-rial as forms, checks, reeelpta,and bills.Copy data from one record toanother, fill in report forms; andtype all work from rough draft

for corrected copy.Interview members of household

to obtain such information asage, occupation, and number ofchildren, to be used as data forsurveys or economic studies.Guide people on tours throughhistorical or public buildings, -describing such features as size,value, and points of interest

Comprehension and expression ofa level to-

-Learn job duties from oralinstructions or demonstration.

Write identifying Information,such as name and address ofcustomer, weight, number, ortype of product, on tags orslips.

Requeat orally, or in writing,such supplies as linen, soap, orwork materials.

Lioniplet ot "ptInciples of tat ionol systems' am Bookkeeping. Internal combustion engines-elect:1c wain, naval. boast!building, nursing, tart mannement. amp Wang

Source: A Supplement to the Dictionary of Occupational Titles,p. A-6.

324

-r

307

TABLE A.3

SPECIFIC ./OCATIONAL PREPARATION SCALE

9 - Over 10 years

8 - 4 to 10 years

7 - 2 to 4 years

6 - 1 to.2 years

5 -.6 months to 1 year

4 - 3 to 6 months

3 - 30 days to 3 months

2 - Short Demonstration - 30 days

1 - Short Demonstration only

Source: A Supplement to the Dictionary of Occupational Titles, p. A-7

to perform physical tasks such as ,climbing and balancing, stooping,

kneeling, talking, hearing, and seeing. We restricted our measure to

physical strength requirements using the DOT scale reported in

Table A.4.

TABLE A.4

THE PHYSICAL DEMANDS SCALE FOR STRENGTHStrength (Lift, Carry, Push, Pull)

Maximum Lift Frequent Lift/Carr

-1 - Sedentary (S)

2 - Light (L)

3 - Medium (M)

4 - Heavy (H)

5 - Very Heavy (V)

10 lbs.

20 lbs.

50 lbs.

100 lbs.

Over 100 lbs.

Up to 10 lbs.

Up to 25 lbs.

Up to 50 lbs.

Over 50 lbs.

Source: A Supplement to the Dictionary of Occupational Titles, p. A-7

325

308

Obviously this scale is insufficient as a complete measure of physical

demands on a given job. Any other measure, however, would, have taken

a prohibitively long period of time to compute.

Our measure of "negative" working conditions also leaves a good

deal to be desired, but again-it-was the only measure that could be

computed without resorting to a full investigation of each occupation

in detail. Each occupation can be rated in terms of whether the job

subjects the worker to certain working conditions including "extremes

of cold plus temperature changes," "extremes of heat," "wet and

humidity," "noise and vibration," "hazards," and "fumes, odors, toxic

Conditions, dust, or poor ventilation." For each occupation, we took

-the number of negative. working conditions as our measure for this

variable. This is surely a weak measure for it fails to account for

the intensity or degree of the negative conditions. A given job

which subjects a-worker to great hazard to life and limb may have only

one negative trait "hazards") while another job might sabjcct :

the worker to some humidity, noise and fumes and thus be counted as

having three negative traits But again we felt this measure was

better than no measure at all.

The problem with using the POT to obtain measures of GED, SVP,

physical demands, and working conditions is that the more than 4,000

specific job titles were not at the time of our research readily

transferable into the three hundred odd census occupation codes.

(Since completing our research the BLS has issued a Census Code -DOT-

conversion table..) It was thus necessary to develop our own conversion-

326_

309

routin- so that DOT characteristics could be assigned to each census

occup tion code. This was accomplished by matching,DOT titles and

six digit codes directly to the list of census occupation names, using

o r knowledge of the occupation st-ucture to make the match. This

rocess took literally months to complete and once done was checked

independently by a group of manpower experts associated with the

Institute of Labor and Industrial RelationsResearch Division at th-e

University of Michigan. Finally we were able to check our conversion

routine against a preliminary version of the BLS conversion table

and make final corrections. It is possible that even after this

painstaking task there are some errors in our conversion system, but

we have a high degree of Confidence in our results. One possible

problem arises in weighting the individual jobs in the broader

occupation codes. Again we had to use our knowledge of the relative

numbers of each specific job in an occupation so as to get an

estimated weighted average GED and SVP score for each census occupation.

In most cases this posed little problem for there was little variance

in GED and SVP scores across specific jobs within aQcensus occupation

code.

After each of the individual census occupations had been assigned

GED; SVP, 'physical demands, and working conditions scores, it was

necessary to group occupations'into individual "Occupation Levels."

This was done on the basis of narrow GED and SVP ranges. The result

was a 17 level breakdown which isreported in Table A.5.

310

TABLE A.5

"OCCUPATION LEVELS "1 BASED ON GED AND SVP SCORES

OccupationLevel GED Range SVP Ranp

1 1.5-2.4 1.8-2.4

2 1.5-2.4 2.5-3.4

3 1.5-2.4 3.5-4.8

4 2.5-3.4 2.0-2.4

5 2.5-3.4 2.5-3.4

6 2.5-3.4 3.5-4.4

7 2.5-3.4 4.5-5.4

8 2.5-3.4 5.5-6.4

9 2.5173.4' 6.5-8.0

10 3.5-4.4 3,0-4.4

11 3.5-4.4 4.5-5.4

12 3.5-4.4 5.5-6.4

13 3.5-4.4 . 6.5-7.4

14 3.5-4.4 7.5-8.0

15 4.5-5.4 6.0-7:4

16 4.5-5.4 7.5-8.0

17 5.5-6.0 7.0-8.2

Unfortunately even with nearly 14,000 cases in the SE0 sample,

this`stradfication by occupation level left individual cells often

too small for statistical purposes after further stratifying by race

and sex. Therefore it was necessary to reconstitute the 17 levels

into a number of "occupation strata;" This was the final step in

preparing the occupation data for linkage to the SEO file. Thp

final five occupation strata were grouped so that each individual

stratum had the same GED range (with the exception of group 15-17)

-and a broader SVP range. The proviso was that each Stratum have

3 28

311

sufficiently sized cells to allnw fo: extensive stratification and

yet yield statistically significant results. The five occupation

strata used in the analysis include levels 1-3, 5, 6-9, 12-14, and

15-17. OccupatiOn level 4 was e)o..litd9d because of singll cell size.

Occupation level 10 was excluded because it was totally dominated by

a very poorly measured specifi job, namely clerical and kindred

workers, nec. Occupation 11 was excluded from the final analysis for

the same reason, its dominant specific job Werag salesmen and sales

clerks, fec. We were so unsure of the CED and SVP scores to attach

to these broad occupation groups that we chose instead to leave rhem

out of the analysis.

The other information collected on census ,)ccupations included

median education, median age, percent black male, black'sfemale,

and white female employment. All of this came from the appropriate

1960 Census volume. It was added to the occupation file and appears

along with all of.the other occupation data in Table A.7. The list

of Census occupations organixei by occupation level with theirr\SE0

code numbs is found in

329

312

TABLE A.6

CENSUS OCCUPATIONS BY "OCCUPATION LEVEL"(SE0 OCCUPATION CODES)

Occupation Level 1

11974 laborers (nec)

07671 manufacturing graders and sorters

04324 messengers and office boys

07654 food, nut, and vegetable graders and packers

07693 packgrs and wrappers08803 laundresses, private household09813 attendants, recreation and amusement

09820 bootblacks09823 chambermaid, ad maids

09824 charwomen an cleaners

0983i elevator operators

09874 recreation and amusement'ushers

11960 carpenters' helpers

11964 gardeners and groundkeepers11973 warehousemen (nec)

10902 farm laborers, liege workers

09841 porters

Occupation Level 211972 truck drivers helpers

07674 laundry and dry cleaning operators

07653 metal filers, grinders, and polishers

07673 textiie'knitteis, loopers, and,toppers

07692 oilers and greasers (excluding auto)

07710 textile spinners,.

09830 counter and fountain workers09835 kitchen-workers (nec)

11971 teamsters

09890 service workes (nec)

Q7694 painter,s

11963 garage laborers, car washers and greasers07695' photographic process workers07632 auto service and parking attendants

Occupation Level 311970 lumbermen, raftsmen, wood, choppers

06535 upholsterers '

Occupation Level 404315 cashiers

08801 baby81,tteri, private household

330

313

TABLE A.6 (Continued)

Occupation, Level 5

07643 .ma ufacturer checkers, examiners09851 gu rds, watchmen, doorkeepers11965 lo gshoremen, stevedores04351 telegraph messengers07730 operatives and kindred workers (nec):06444 log and lumber scalers and graders09834 janitors and sextons09875 waiters and waitresses04304 transportation baggagemen04323 mail carriers04353 telephone operators64360 typists

05383 hucksters and peddlers07714 cab drivers and chauffeurs09815 bartenders04313 bill and account collectors07631 assemblers04320 file clerks

Occupation Level 607685 mine operatives and laborers (nec)09812 attendants, professional and personal service (nec)06494 construction and maintenance painters09810 attendants, hospital and other institutions09854 sheriffs and bailiffs04350 stock clerks and storekeepers06435 heat treaters and annealers07640 railroad brakemen07690 motormen--mine, factory, etc.07704 sawyers07705 manufacturing sewers and stitchers07712 stationary firemen07715 truck and tractor drivers09842 practical nurses04354 ticket, station, and express agents04343 shipping and receiving clerks

Occupation Level 7

07703 sailors and deckhands06415 cranemen, derrickmen, hoistmen09853 policemen and detectives07720 textile weavers

06425 road machinery operators04341 receptionists

331

;314

'TABLE A.6 (Continued)

04345 stenographers04352 telegraph operators06404 bookhinders06431 forgdmen and hammermen07641 bus drivers07645 bus and streetcar conducors07651 dressmakers and seamstresses07670 furnacemen, smeltermen4 pourers07691 motormen--street, subpay, el, RR'07713 RR switchmen09814 .barbers09843 hairdressers and Osmetologists /

Occupation Level 8:06513 metal rollers and roll hands07721 welders and flame-cutters06413 cement and concrete finishers06434' glaziers07630 asbestos and insulation workers07672 metal heaters07675 meat cutters09850 firemen, fire protection10901 farm foremen06512 printing pressmen and plate fitters01015 athletes06490 millers, grain, flour, feed, etc.

UCcupatiGn Level 907613 appren. building trades (nec)06401 bakers

06495 construction and maintenance painters

06501 paperhangers06505 plasterers06514 roofers and slaters06515 shoemakers and repairers

07652 dyers07602 -appren. bricklayers and masons06503 photoengravers and lithographers06411 carpenters

Occupation Level 1004312 cashiers

04361 clerical and kindred workers, nec.04302 library attendants and assistants04340 postal clerks

315

TABLE A.6 (Continued)

Occupation Level 1104325 office machine operators07650 deliverymen and routemen04305 banktellers04310 bookkeepers04333 payroll and timekeeping clerks06420 decorators and window dressers05396 salesmen and sales clerks (nec)

Occupation Level 1202212 farmers (owners and tenants)06492. metal molders01180 sports instructors and officials09825 cooks (except private hOusehold)01185 medical and dental technicians03252 RR conductors04314 vehicle dispatchers and starters04342 secretaries05380 advertising agents and salesmen05393 real estate agents and brokers06402 blacksmiths06410 cabinetmakers06452 inspectors04303 doctors and dentists att Adants

Occupation Level 1306545 craftsmen and kindred workers (nec)07634 blasters and powdermen06480 mechanics and repairmen (nec)06453 linemen and servicemen06523 structural metal workers06424. engravers01193 therapists and healers (nec)03254 floormen and floor managers, store06403 boilermakers01161 photographers02222 farm managers03250 buyers and department heads, store03265 ship officers, pilots, pursers03280 postmasters03285 purchasing agents and buyers (nec)06421 electricians

06454 locomotive engineers06460 locomotive fireman06461 loom fixers

333

316

TABLE A.6 (Continued)

06471 airplane mechanics and repairmen06472 automobile mechanics and repairmen06473 office machine mechanics and repairmen0647'4 radio and TV mechanics and repairmen06475 RR and car shop mechanics and repairmen -06491 millwrights.06510 plumbers and pipe fitters06520 stationary engineers

06521 stone cutters and stone carvers06524 tailors and tailoresses06525 tinsmiths and coppersmiths06530 toolmakers, die makers setters

07604 apprentice electricians

07605 apprentice toolmakers and machinists07614, apprentice metalworking trades (nec)07615 apprentice printing trades07701 power station operators06430 foreMen06465 machinists06502 pattern and model makers06405 brickmason and stonemasons

Occupational Level 1406423 electrotypers and stereotypers06451 jewelers and watchmakers06470 mechanics and repairmen07610 apprentice mechanics, except auto06414 compositors and-typesetters

Occupational Level 1503260 publiC administration inspectors01074 draftsmen01104 funeral directors and embalmers01164 radio operators01190 electrical and electronic technician01191 other technicians01192 technicians (nec)

04301 agents (nec)01072 designers01154 personnel and labor relations workers01182 teachers--elementary schools

05395 stock and bond salesmen07642 surveying chainmen, rodmen, axmen01102 farm and home management advisors01111 librarians

334

317

TABLE A.6 (Continued)

01150 professional nurses01160 pharmacists01163 public relations men01174 statiticians and actuaries01181 surveyors01183 teachers--secondary schools01184 teachers (nec)

03251 buyers and shippers, farm products

03262 building managers and supeiintendents01090 metallurgical engineers01084 ,industrial engineers01093 engineers (nec)'01012' airplane pilots and navigators05385 insurance agents and brokers

Occupation Level 16

01014 artists and art teachers01073 dietician and nutritionists01171 social and welfare workers01001 accountants and auditors01120 musicians and music teachers03253 credit men03275 qficial--Lodge, Society, union04321 insurance investigators01085 mechanical engineers01082 /civil engineers01170 /religious workers010911 mining engineers01080/ aeronautical engineers01075) editors and reporters01085 electrical engineers01165. recreation and group workers

Occupation Level 1701020 authors

03270 public administration officials--administrators01173 psychologists

01030 college presidents and deans

01103 foresters,and-conservationists01013 architects01021 chemists01023 clergymen

01031 faculty--agricultural sciences

01032 faculty--biological sciences

335

318

TABLE A.6 (Continued)

01034 faculty--chemistry01035 faculty--pconomics01040 faculty--engineering01041 faculty--geology and geophysics01042' faculty--mathematics01043 faculty--medical sciencesr1045 faculty--physics.01050 faculty--psychology01053 faculty--social sciences,.(nec)01054 faculty -- non - scientific subjects

01060 .faculty--subject not specified01071 dentists01081...chemical engineers01105 lawyers and judgesb1130 agricultural scientists401131 biological scientists01134 geologists and geophysicists01135 mathematicians01140 physicists01145 miss. natural scientists01162 physicians and surgeons01194 veterinarians01172 economists

3C6

._7

319

The columns in Table A.7 refer to the following:

Column Datum

1 SEO Cens-Js Occupation Code ,

2 Physical Demands

3 Number of Negative Working Condition Traits

4 GED SCORE

5. SVP SCORE

6 Median Years of Schooling - -Males

7 Median Years of Schooling -- Females

8 Median Age--Males ,)

9 Median Age -- Females

10 '. Imputed "Occupation Level"

11 Percent Black Male Employment

12 Percent White Female Employment

13 Percent Black Female Employment

3,

/

...

SCIri _JO

9'0 601.'0 0'0 'LI. 5Er 10E S91 ZLL 04 09 0 PC,0C 5E0'0 010'0 'Ll S'EE 7'hf 591 7-L1 0.9 0'9 '0 720.0 11770 0E0'0 'L1. S'6Z E E 1791 LL o8 0.9 0 0'0 0E0'0 100'0 'Ll S'Ef 5SE 5'91 91 3"11 ic/ '0 910'G 057'0 SZO0 'Ll LZI: FnE 5291 9'91 OR 0'9- '0 0'0 190'0 OZ00 'Ll S'ff 6'Pf 5'91 9*9I 34 0 070'0 7950 PZO '0 *91 Z '9E Pbi f 0.0 O'S '0 61'0'0 018'0 900'0 'St 01E 7'91 L'91 C'S '0 00'0 1E00 100"0 'Ll L SP fct 0%1 PL 0.0 0'9 '0 013'0 150'0 1.00*0 '51 9'£S P'Zh L71 "r1 0'L S't) 7 1.00*(1 LZ00 P00'0 *Li 4'61 '71 6'L 3'9 '0 stip° Ent*o 6E0'0 '51 6'9f 7'0E Sql n'at O'L O'S 0

........ I. *** 0'0 600'0 L00'0 'St CAE l'ff ''II (al 7'1 7'S 'C

o'n zoo.° r00'0 'at L'6E 6'9E E'r1 5'00 OP 7'S '7

0'0 SI0'0 LOO*0 'Si L.60 0.0( 6E0 c.qi ht. U'S '0 9'O £00'0 100'0 '91 L'6E A9E 6'Et 1'11 R'L I'S .'0 0'0 1'70'0 £00'0 'St col 9'Zt rst f'L t'S '3 0'0 600'0 500'0 '91 9'hr E'9E 7'S1 7'91 P'L h'S '0 0'0 h00'0 80C'O '91 L6E EOh 6'E0 1'01 0'9 I'S '0

P*0 1110'0 100'0 *L1 L'6E 0.0E 6'EL /91 C'R 0'9 '0 O'C' StO'C 900'0 '01 L'6F 6'SE 6"'EL (-01 L'L r'S '0 C09'0 E91''0 S00'0 '91 S'6E 17'6C 9ht 0'91 S'L JOS '0 100'0 LS0'0 010'0 '151 L'ec 6'71 6'71 O'L S'h '0 971'0 6PL*0 LI0*0 '01 l'hh 69£ 7'Et L'71 S'L i'n '0 cod'o PLI0 900'0 *S1 S'th ut% n'ct c'Et t't 'o

1.01'0 070'0 920'0 'LL 5' #01 6',07 0'91 rLt 0'11 0'9 '0 910'C FIGL '0 800'0 'SI 6'87 t'67 S71. OL S' '0 LoO 6I2*0 110'0 L'1,/ A'PE P'Ll PL 0P 0'9 '0 rice 767'0 910'0 'L1 9*th E07 fr'Ll 11-LL 0'9 0'9 '0

0 zn0 6L70 070'0 'L1. 991 P'017 h'LL PLl 0'8 0'9 '0 00 M*0 0'0 'LL P'LL tLt 0'0 0'0 'C 0'O 780'0 0'0 'Ll a. n'Ll t'41 0'0 0'9 '0 "600'0 hOZ'O 010'0 'Ll O'Lf h'Lt h'L.1 O'R 0'9 '0

r1'0'0 9E0'0 E10'0 'Ll "'11 '09 P'a '0 L00.0 191'0 070'0 'LC l'Oh S'Oh h'Ll h'Lt 0'9 3'a '0

E10'0 int.° St0'0 'Ll L'rh 10tE h'Lt t'LI 0'9 C'0 '0 0'0 .970'0 Et0'0 'LI E'9£ h'LI h'Ll 0'9 9'9 '0

700'O OZO*0 ht0'0 'LI ** '"QE h'Ll 0'0 0'1 '0' 900'0 190'0 710'0 'LI **

' b-nt, hLI 1,11 0'9 3'9 -e cloco two 'Av.° -Lt *** hLt btt n'n O.4 '0 90/-0 551'0 'ITO '0 'LI 6'"f Z t'Ll hLi 3'9 ''9 '0 O'C 900'0 8E0'0 'Ll 07 P'LL bLL C'9 '0 c000 917'0 6E0'0 hf,h LLn nL hLt 7'9 9s '0 coo 610'0 L90'0 LL Vim 7.11 P71 I'LL nR 0-11 '0 0'0 160'0 4 (00 '0 'St E'rIS r'Sh 7'S1 511 3'L 0'S '0

E 9L00 6 1.0-0 L1. LEE O'LE 5'91 0'9l 0'11 0'9 '0 P1000 GLZ0 L 0'0 'LL Cm, 6PE 0'91 9251. ;L S'S 0 5000 SLO0 PS 0 '0 '147 5'71 5'S E'E '0

900'0 I1E'0 11 '0 '91 E'Rr 0tr l'St CFI S'L S'h '0

0'0 120.0 sno 6 -LI e.0'°1 O'S 1.00C 500'0 £00' *S1 ** .'9r * S'rt 3'L 0'S '7 h1VP 917f'0 E10'0 'St l're "RE t'El l'hl O'L O'S '0 E00'C Z91'0 SO0'0 '11 9'7h 0Ch R'71 r'r1 0'9 0'S 'n

11,0,4 11. is -

t

'1

7 'I '2

t/ '7

'2

'Z

'2

'1

'7 '

'7 'I,

'7

'1

'1

'2

'1

i '7 '2 '2 7 '2

'7 'Z

'7

'7

'7

'Z

'Z

-z

z 7 '7 '1

f -

'1

'E '7

't

'7

'7 '1

c^11, 011 cill "CIL 1E11

OF 1.

071 lilt S011 hOlt E011 2011

01. E601 7601 L601 0601 .

SP31 1000 E1131. ZPOL 1P0( 0001 SLOI tLOL fL31 7L31 1101 0101 1901 hS01 ES01 2501 1501 OSOl ShOt EhOt ZhOt 0101 OhOl. SrOt qrot. 7E00 1E01, OfOti (7011 77014 1201, orolt slot'

11011 71011 01011

10011, S6601 0000£

(ET) (zT) (TT) Cot (6) (8) (t) (9) (c) ('7) (E) (z) (I)

aaop oas a ma NoLwan000

auvI

OZE

feS*0 900'0 '11 uer rah 7f0'0 701.'O 110'0 It 9'Of 1'97 h10'0 AS1'0 571'0 t'61 Cf7 700'0 610'0 101'0 'S 7'fb C'Ob rove 1710 b00'0 '91 C'Sf e'bf SSO*0 (6L'O 910'0 'S 0'67 h'LZ 900'0 070'0 1700'0 'h J'/1/

110'0 001'0 170'0 *71 7'011 S'Ob oto.o 081'0 610'0 '5 0'7h R'LE 1170'0 RSV() L00*0 '01 1.'9C t'97 LOCO 070'0 700'0 'ft 0'6C 6'PE 700'0 909'0 f00'0 '11 L'7t 9'0C /100'0 610'0 701'0 'S S'7b 670'0 tC6'0 £00'0 '71 L'7( 7'lf 5'410'0 LOCO '610'0- '01 l'lf 1'77 500'0 171.1'0 110'0 '51 h5W urn oe.s. * 700'0 760'0 SO0'0 'ft R.Th. 7.7h 100'0 70r0 100'0 'Et O'TS 5'05 100'0 660'0 1'70'0 '91 A'91b L'Lh 900'0 Sel'O 010'0 'Ll 9'Lb f'Ph 100'0 (00'0 1.00.0 11 L'Sh 010'0 19C0 100'0 'St 0'9S L'hy S00'0 7110'0 PlO*0 'St ALII h'sh 170'0 7.Lh'0 P10'0 'ft L'7.b O'C 6(1'0 b00'0 '91 6'7h O'AE 0'0 f00'0 S00'0 '71 rtS 100'0 L10'0 000'0 'St 'Rh 700'0 Lt7'0 b00'0 'ft C'S4 6'011

700'0 1E0'0 hZ0'0 'Cl S'fh 900'0 Oh0'0 (90'0 '71 L'tS 7'Fb

0'0 1110'0 L10'0 'Li 7'6( h700 505*0 1E0'0 'Et L'14( f'7h 510'0 077'0 Of0'0 '51 6'51 6'7C 1100'0 771'0 1110'0 'St E'9( L'7C f00'0 9b0'0 616'0 'St 6'9F /'Of 7h0'4 SLYO 1E0'0 '71 flf, Vhf LF0'0 ofvo

1LS*0 hrvo

010'0 F70'0

'51

'St

Vrt, 11'14

b'LE 6'4C

LLO*0 LLCO ht0'0 '51 S'bh h'hE 100'0 AC0'0 100'0 'St C'L( 41f 910'0 00(0 Sf0'0 '71 A'tf O'ft 500'0 197'0 ffO'O 'Ll 9*Ch 7'011

LOO*0 b0E0 b10'0 'St 9'0h ('RC 600'0 00i'0 710'0 'Ll 6'61 h'cir

TO0'0 1E1'0 010*0 'Ll L'6( 9'6f Fr9'0 145'0 100'0 '91 Chn C.'PE

S10'0 665'0 710'0 '91 S'Sh CLF Ab0*0 SLUO h11O*0 '91 fl'ff

1100'0 A60'0 110'0 'St rot f'5E b00'0 ((7.0 900'0 '51 L'9C fqh 701'0 190'0 070'0 'Ll 9'1h l'fb Ervo 911'0 610°0 'ft q'Ob Z'hr 700'0 R[0'0 L10'0 '51 1'6C b'Sh 400'0 ("67'0 1400'0 "St 6'lb 7'lb 500.0 -(01'0 500*0 Aph E03'0 Oh0'0 100'0 'St 4'1h

i b50'0 1716'0 [00'0 '51 l'6( 50h

62C

17.71 (-.?t 6's 0.11 *0 .1 fret's'

h'71 S'h S'C '0 '7 czfnr .A11 901 0'7 3'7 '0 '7 h/f0C 7'71 f'71 irr *0 't mho 9'71 4'41 O'R l'S '0 1 17111(

Z*71 h'71 O'f O'C '0 '7 07f110

+7'71 hE '0 'h S1f110

0* l irn 0'7 irtt .0 .1 .httho 1Z1 f'71 r.f .0 '7 11010 0'7 3'71 O'f S'C *0 1 ZlfbO Sfl L'71 O'S O'b '0 1 01fb0 '5'71 L'71 O'S O'h '0 '7 SO(b0

s'Ol O'C Q. '0 'C hOrb0 S71 S'71 0'9 7'h 1 *7 10fh0 lfl O'h O'h '0 'f 701110

s'n 1'71 O'L 5'h '0 '7 lOCh0 067C0

ci'n 71 O'L 0'4 '0 't 5117E0 E'71 S*71 O'L O'h '0 1 047f0 S'M 1'71 O'R O'S '0 't SLZFO L*71 6'71 S'L 5'S '0 1 °UFO

S'Ot O'L O'h 1 '7 So/CO ('II h'6 O'L O'S '0 'Z. r97f0 h'71 t'71 5'4 S'h 097C0 S'll 0'71 S'I O'h '0 '1 hS7C0 S71 rrt 0'f1 VS .0 1 r57r0

5'01 0'9 O'h 1 T. 717(0 0'11 O'L O'S '0 '7 1S7f0

h'71 C'L O'h '0 q OSZCO O't' O'h '0 '7 77770

"0 CA S'S S'f '0 71770 S6110

'Ll O'A 't '7 66110 1'91 h'911 S'9 O'h '0 '7 F611C /71 0'71 0*1. S'h 'C '7 76110 9'71 P*71 O'L 5'h '0 '7 16110 C'71 1'71 C'L S'b '0 '7 06110 7Et 0'71 0'9 O'h 0 '7 S911( 1'91 h'91 3'1. O'S '0 '7 hell( 9'91 7'Ll O'L 0'5 '0 '7 fAll( bqj o'Ll C*9 O'S '0 '7 7.011(

cri O'L Vs *0 '7 10111

S'91 5'91 0'9 SF '0 'E 0811( 5'91 O'Ll 0"9 '0 *1 Silt( O'ft 1'91 3'L O'S '0 1 ht.C11

h'il S'il P't P'S '0 't tLtIl 6'51 A'91 O'L 0'9 '0 't 71,111 q'qt coqt S'L 0'5 '0 1 11.111 toft 9'L F'S '0 ntAtt L'hl 1'51 S'L b'S '0 1 59111 UZI 9'71 O'L S'h '0 '1 11911.

b'St 1'91 O'L O'S '0 1 09111 b'Ll 5*/1 'O'R 0'9 '6 '7 7911- b't 4o71 O'L 0'4 '0 '7 1911" 7'41 7'01 O'L O'S '0 .7 09111 t'fl 6'51 S'9 0'5 'n 't bqtt

r'Lt 3'9 3'9 '0 'f ES11. C'Ll O'L O'S '0 1

* 1511 ZEt Cft O'L 3'5 '0 '7 OSIl

(cT) (zT) (TT) (9T) (6)' (8) (L) (9) (S) 07) (0 (z) (T)

(panuT4uo0) L 'V TIUVI

TZE

ZOO*0 SZO*0 777'0 Z1 0'0 100'0 610'0 'EL

0'0 SO0'0 S90'0 'A ZOO*0 510*0 150'0 o 'FL 0'0 100'0 Ch0'0 '11

700'0 St0'0 7h0'0 'Et

100'0 U0'0 710'0 -rt ioire rovy 940'0 'EL

100'0 1110'0 L70'0 'EL 100'0 100'0 170'0 'hi 100'0 C10.0 SZO*0 'EL LoVo novo zovo -11

040 C00'0 rZ0*0 'CI 0'0 700'0 Z00'0 'FL

tO0'0 610'0 110'0 'CI

t00'0 POO*0 tIO*0 '7L

600'0 090'0 010'0 'hi

600'0 1100'0 CP0'0 'S

200*0 L10*0 690'0 '9

n'D 770'0 710'0 'R

0:0 two L70'0 'CL

r00'0 8E0'0 190'0 'L

700'P 990'0 410'0 'EL

0'0 h00'0 LC0'0 .L

600'0 h111'0 120*0 'ft

0'0 L00*0 000'0 'hi 0'0 L00*0 010'0 'CI RZ0'0 P7h0 rZO'n 'IL

LC0'0 000'0 L60'0 't

hoo'n r10'0 07C*0 'ht 0'0 700'0 6LZ0 '8 0'0 1O0'0 ShO*0 '6

100'0 710'0 970'0 '7I

tovo row° IWO '1%

own 715'0 670'0 L l'i'0 100'0 .9Z0'0 'CI 0'0 900'0 670'0 '71

117.0'0 0E1'0 790'0 '6

600'0 R6E*0 600'0 'LL *0'0 090'0 S00'0 'St 900'0 P77'0 1'10'0 '71

100'0 ZhO*0 th0'0 'h 900'0 560'.0 1110'0 'St 070'0 P4S.0 920'0 'S L60'0 616'0 ZOO*0 :S

0"-0 CZO*0 0'0 'R

100'0 O51*0 h00'0 'ZL

ZZO*0 595'0. (t0'0 '01

SSO*0 SaR*0 h00'0 'S

100'0 617'0 900'0 '9

C70'0 IC6'0 Z00'0 'S

100'0 SLZ*0 SO0'0 'I_

1110'0 090'0 L91'0 'S Z70'0 471'0 1P0'0 '9 610'0 L76'0 700'0 t 00*0 9L0'0 001'0 'q htn'n 1S6'0 100'0 '7L

S70'0 Sh6°0 700'0 't ewe IWO hht'O 'OL

ckc

CRC n'th r'01 n'n S'S S'C s *r 7c.h901

0'Sh 6'6 O'L 0'4 1 " 16h90 6'Eh .6.P I'9 1E '7 'Z

01 0',01 -7'7h 9'01 hOt 8'9 LE '0 '1'

CIOP:70

S'6C E'70 VAL P'Ot t'L O'h '0 '1'

r4Oh I'll S'L O'h 'C .0

111 O'L0 l'h O'L O'h 7 "tt

7'04 ,'4f. 6'tt I'7t O'L On :0 '1'

_to7t O'L O'h -'0 '7 vrr hClf 901 f6 3'L O'h 'Z 'r

6'111 L'll 9'11 7'71 O'L Oh 't -r

yfh 9 O'L 0'0 'I 't 1qh90 7fC o L'LL O'L n'h 't 'Z

9'6C hhh S'll r'LL S'L O'h '0 'L

6-LE. /*Eh R'Ot 66. L'C L'Z 't 'C

S'4E i-rr t'7.1 1-7t S'9 CPC % 't

h'Lh 6'6 0'9 O'h '0 '1'

ihhtr°0

0Shq0

::: i'tS 4.44...11'6 O'L O'h 'L '7

t'bh 11'6 O'h O'C 'C 'C Srh90 11E100 L'SC .9'11t 0'9 vr -(1 *C

605 0'01 O'L n'h '0 'L 7rh90 c'f h f'11 O'S O'r 'Z. '11 tr490

9'hh L'Ilh 6'01 S'LL rt. O'h 'Z 'Z Ei L'Oh P6C O'll 0'71 O'L R'C '0 '7

6'6E t'f, 6'0 O'E 't 'E

nth U O'P S'C 'Z '1'

S'Lh O'Lh O'Lt R'11 O'L O'h 't '1

S'th S'RC h'Zt fIt O'S O'h '0 :r g * 4b'th ..6n S'h O'C 't 'I

S'Oh r on 7:71 t'71 O'R O'h '0 'Z

47'0h 9'11 3'9 O'E '7 'I Cif L'qr L'rh 0U1 66 0'1 O'f 'Z 'f

tlhlg

O'fh 0'01 3'9 O'h '0 '1'

4 O'PC W6 C'L YE' 'Z 'C

:::::

O'Ch L'6C s-tt rzt vs 7'0 '0 '7

S'Sh UZI, L'4 Z6 e'L 'O'E 'l 'f t'LS S'P D'q 3'h 'Z -t,

i°!;(9!

ionno

hqh h'RE 6'6 POI 3'S O'r '0 '7 I'Sh .4e.,..9.6 S'9 O'h 'Z 'h

LLh Lhh 8'7L S'hl S'9 O'S '0 '7 snrso L'811 h'OS 9'71 L'Z1 0'9 0'0 '0 'Z r6CSO L'LZ L'St 0'01 0'6 O'Z O'C '0 '7 06CS0 q'Ch or, 9'71 1.71 0'9 L'h '0 't SACS° E14 6'Sh l'T,t Z'OL 0:E O'C '0 'C CPESO 8'8C C'LC t.'It 0'71 O'C O'C '0 '7

O'Lh smt9'11 0'9 1 nr '0 'Z = L'6C 0'61 8'71 r'Et 0'9 O'h '0 'Z oorso Z'RE P'11C t'ZI hZt 7'4 O'C '0 'L

L'SE O'LC 7.'71 E71 or O'C '0 't

R*67 O'Oh CIL ('It O'h I'C '0 "Z

9'hh O'Oh UZI 7*Z1 O'S n'r '0 'L

R'01 h'Ot Of S'Z '0 'Z

1(1111!!!!

ISOM

0'6Z 9'RZ S'Zt S'7t O'r O'C O 't

E'6E 9'hC 0'71 11'11 O'h vr '0 'L

ZSCh0

6.7r CRC 9'7t o'71 O's 0'f *0 el

OSCh0 ShrhO

n'fiC L'IC 5'11 Aet S'E O'C '0 'I'

Z'Sh P'Oh h'tt 6'71 0.11 0'4 '0 '7

C7101M nhr vrn 9'71 0'71 vn O'h '0 't

Crr SW 5'71 c'rt OS Of '0 't

(ST) (n) (TT) (OT) (6) (8) (L) (9) (g) (h) (£) (Z)

(PanuT4u0b) L'v

ZZC1

CO

323

/TABLE A.7 (Continued

(1) (2) (3) (4) ,(5) (6) (7) (8) (9) (10) (11) (12) (13)

06'493 2. 2. 4.0 6.0 10.5 47.1 12. 0.030 0.022 0.001

04494 1. 0. 2.8 3.4 12.0 11.4 39.9 40.3 6. 0.014 0.145 0.007

06495 2. 1. S.0 7.J 9.1 11.7 45.5 41.9 4. 0.068 0.018 0.002

J4501 ' J. 1. 3.0 7.0 0.6 d.a 51.3 47.9 9. 0.108 0.126 0.002

0659 J. 1. 4.0 7.4 12.0 r 42.2 13. 0.007 0.016 0.001

ob5p3 2. 0. J.c 7.2 12.2 12.0 18.8 19.2 9. 0.011 0.051 0.001

04504 1. 0. 4.0 7.0 11.1 50.7 13. 0.012 0.032 0.0

06505 4. 2. 3.0 7.0 9. 40.5 9. 0.166 0.003 0.001

Co510 4. 1. 4.0 7.0 10.2 10.5 44.6 44.7 13. 0.032 0.003 0.0

(3512 J. 1. J.2 4.2 11.5 11.1 38.0 J8.7 8, 0.01d 0.019 0.003

)651J J. J. J.0 5.5 1.2 9.d 42.9 40.0 8. 0.118 0.034 0.003

065 fu J. 1. 3.0 7.0 u.) 34.9 9. 0.099 0.001 0.0

0651!, 2. 0. 3.0 7.0 u.4 S.5 51.5 48.3 9. 0.132 0.036 0.004

04520 J. 3. 4.0 7.0 12.2 12.5 44.0 42.3 13. 0.017 0.006 0.0

04521 3. 3. 4.0 7.0 J.J 45.6 13. 0.042 0.012 0.0

0452.1 J. 2. 3.6 6.5 10.0 41.2 1,3. 0.037 0.005 0.0

04524 2. 1. 4.0 7.0 d.4 8.9 57.9 51.4 13. 0.35d 0.194 0.014

06525 J. 2. 4.0 7.0 10.d 11.0 39.4 42.6 13. 0.014 0.011 0.001

04530 3. U. 4.0 7.0 11.7 11.4 42.3 41.6 13. 0.009 0.007 0.0

06515 2. 0. 2.0 3.5 9.7 Y.4 39.4 42.0 3. 0.044 0.099 0.006

06-545 2. 1. 3.5 7.0 10.2 10.3 39.4 40.5 13. 0.056 0.01d 0.001

06555 O r4o O ***

07401 J. 2. 4.0 7.0 11.1 19.7 13. 0.023 0.023 0.0

07b02 J. 1. 3.0 8.0 11. 23.0 9. 0.107 0.004 0.0

-0760.1 J. 2. 4.0 7.0 14.0 21.7 1J. 0.01d 0.0 0.0

07604 1. 1. 4.0 7.0 1.4 2.1./ 1J. 0.009 0.013 0.0

07401 J. 2. 4.0 7.0 12.1 23.1 13. 0.017 0.019 0.0

07410 4. J. 4.0 '7.5 12.2 23. 14. 0.011 0.0.16 0.0

07412 " 4. 2. 4.0 11.5 12.2 23.4 14. 0.007 0.0 0.0

07613 J. 1. 3.0 6.5 10. 21.6 9. 0.047 0.029 0.007

07614 4. 1. A.0 7.0 12.2 23.5 13. 0.025 0.007 0.0

07415 2. 1. 4.0 7.0 12.2 22.6 13. 0.018 0.034 0.0.

0701007621

..Sus

07430 2. ' J. 3.0 6.C\ 10.4 ***** 37.7 U. 0.022 0.022 0.003

407411 J. U. 2.5 3.0 10.J 10.4 30.4 J9.1 5. 0.040 0.424 ).02J

076.12 3. 0. 2.4 2.4 10.5 10.5 2S.J 37.6 2. 0.092 0.019 0.002

07634 4. 3. 3.5 7.0 8.7 42.0 1J. 0.117 0.009 0.0

07645 2. 0. 3.0 5.0 6.9 43.7 7. 0.026 0.00d 0.0

07640 J. 2. 3.0 4.0 10.d 41.0 6. 0.031 0.001 0.0

07641 3._ 1. 3.0 5.0 9.d 11.7 44.9 40.6 7. 0.090 0.092 0.005

071142, 2. 0. 5.0 6.8 11,7 25.1 15. 0.044 0.025 0.002

07643 2. 0. 2.5 2.5 11.6 10.2 40.5 41.4 5. 0.019 0.437 0.019

07445 J. 1. 3.0 5.0 10.2 47.0 7. 0.254 0.009 0.0

07650 J. 0. 4.8 4.d 11.0 11.1 33.4 38.7 11. 0.080 0.024 0.002

07E51 2. 0. 3.0 5.0 0.8 9.5 50.5 55.3 7. 0.004 0.079 0.075

07652 J. 3. ,3.0 7.0 d.j.. ***** 3/.6 9. 0.091 0.023 0.007

07053 3. 1. 2.0 3.0 9.4 10.1 41.7 40.7 2. 0.077 0.05b 0.003

07654 2. 0. 2.0 2.0 d.2 8.6 17.4 41.1 1. 0.033 0.441 0.056

0/470 3. 3. 3.0 5.0 8.9 41.7 7. 0.240 0.015 0.001

07671 2. 0. 1.8 2.0 8.9 9.1 40.4 42.7 1. 0.035 0.611 0.077

07672 J. 3. 3.0 6.0 8.9 45.5 s1/ 8. 0.095 0.010 0.005

076/1 2. 0. 2.0 3.0 9.0 9.1 37.5 36.6 2. 0.007 0.666 0.011

07474 2. 2. 2.0 2.5 9.4 8.9 40.8 42:3 2. 0.096 0.442 0.262

07675 J. 1. 3.0 4.0 10.7 10.0 41.2 40.5 8. 0.039 0.027 0.004

J7400 2. 0. 4.0 6.0 9.0 54.5 12. 0.009 0.850 0.060

07465 4. 2. 2.5 3.i U.0 S./ 40.5 37.1 6. 0.044 0.004 0.0

07640 2. 0. 3.0 4.3 6.3 43.7 6. 0.106 0.001 0.003

07491 J. 1. 3.0 5.0 9.d 46.0 7. 0.132 0.011 0.005

370',2 J. 0. 2.0 3.0 d.9 41.1 2. 0.077 0.006 0.001

341

Zfe

600'0 ZE0*0 6EZ*0 't Z'PE t'LC 4'6 81 9'1 '74 'h hL6ll

(00'0 C10'0 0f1.0 't Z'OE t'LE 4'01 L'Ot O'Z C'Z .6 *c FL61t

rev() Lovo 6e7'o *r 6'1C o'R 4'7 '4'1 '7 '4 7/611 1'00'0 ZG0'0 SRZ0 'Z 0.0h P'L O'r n.E 0 r 1L611

rove s00.0 1.60 'E S'hf 9'9E f'R 6'L P'h rt '7 'h et6tt 709'0 h00'0

E00'0 Zt0'0

IWO L61'0

'S

'1.

Cht, S'rn S'Fh

'0 6w SR

S'Z O'Z

S'Z O'Z

'0

'0

'4

V S9htt b96tt

910'0 R10'0 FZh'0 'Z ,9'9Z vez 4'6 h'6 O'r. fZ 'F frifitt

700'0 P00'0 760'0 'r G'Zh. o'n or 'h 7. ,1614

700'.0 S00'0 OhZ0 't 8'1E OZ 0'7 'h 09611 0'0 Lt0'0- 170'0 !I. 8'Zh h'Pt l"6 O'S ff 0 1- 50601 Love PAE0 11.0'0 't 8'7h h'Pt I'S 106 3'Z O'Z '0 'E runt PS0'0 440'0 007'0 't t'9F S'L 6'1. 0'7 -o'r '0 'h 70601 0'0 rtvo PE0'0 '8 ,6'Sh 0,1.'6 3'9 n'r "0 '7 10601

960'0 41'1'0 Ghl'O 'Z S'hh 9'LZ L'6 0'01 S'7 7'Z '0 '7 Off60 Z40'0 808'0 1E0'0 '4 A'ff q'tlf cl'Ot L'04 O'f P'7 '0 '7 4060 600'0 P67'0 0E0'0 L.Lt POI vot o'r o'r *0 *r. 4060 040'0 660 57.0*0 'S Vth P9 VII VA VI' 8'7 1 '7 0(1860

200.0 rntro

6h0'0 ;Tv°

SZO*0 huvo

'9 6'9h VP, vfir F'PC

f'!1 l'Zt C'Ft f'Zt

S'E O'f '0

*t

7 ,'r

h436O CC060 74360

rocp 970'0 140'0 '4 r4 (rs vit 69 yr yr 't *7 1Sq60 '100'0 700'0 610'0 'P 'RE 1'71 0'9 O'F '4 'S 04960 0(4'0 eovo um' q'tt vrt crt o'S C'F 0 7 Fh860 hq1*0 66C0 toro '9 S'9k 6'hh t'tt (Li O'h O'f '1 'F Z060 I.10'0 LOCO L'OE hfh f'41 O'Z O'Z '0 'C thP60 154'0 OP0'0 SPO*0 '9 O'h *o* r 0060 191'0 17h'0 901'0 't 9'17 6'P 0'6 of O'Z *7 'C SEPKO

670'0 960'0 6h7'0 '4 r'6h C'04 L'P coR 8'7 4'7 '0 'C bfF160

SLO'0 8W0 11.0'0 'ft f'1; 4.4h L'01 011 O'L 7r060\

t.0 [WO LSt0 't R'RF 914 4'01 4'0 3'7 O'Z '0 'F 10160 ER0'0 or(re P70'0 'Z r'on chz fnt c*nt for 0Z 0 7

1 01'0601

sfltD L0,0 LLO*0 'Zl hth fru, 0'6 h'6 0'9 L'F 'Z 'F (JR6V Oot PrO tern 1 gLE 4'0 6(1 0r 0t 0 I' 1,70601

L64'0 9LE0 010'0 thh h'Zh L'8 4'8 a? z t */: 17060 F90'0 Pt8'0 P00*0 '7I L'ZG 6'94 0*01 7'01 0'9 O'h '0 1' tZ960 h10'O 070'0 194'0 't 671 O'Z O'F. '0 7 ,?,71160

f00*0 E01') cthO*0 S hFh Z'Sh 4'01 1'01 0E 0*E 0 *Z 61R60 010'0 Lt0'0 ZPO*0 'L L'6f 6Ph Ult 7'6 O'S O'f '0 '7 ht060 600'0 811'0 401'0 't USE tail f6 O'Z 0'7 0 7. 11060 hu.-0 6Lh'0 1,10'0 '9 0'61 t'tif 7'11 1!11 IE 3'7 'r rte6o 691'0- h94'0 080'0 '9 h'6f l'ir P'01 9'01 S'E O'E '0 'f 01060 849'0 E670 720'0 '4 l'Sh h'6h hP h8 4'7 S'Z '0 '7 hOR80 p-00 fah0 900'0 't 0-7; clz '0 '7 C08110 .9P7.'0 b600

889'0 bL0*0

h00'0 100'0

'Zt *b

h'CG 8'74 6,,Z 29L

9'P 4'f,

ct*fi

0'9 nZ

O'h 04E

'0

0 '7

*r

70080 10080

870'0 /.97.'0 4/.0'0 ''S 7'04 9*PC 1'6 h'6 f'E 4'7 '0 '7 GfLLO 240*0 ram., PSO*0 ft s.9r hE S'Ot L'6 P*; or 'E 'h 17LL3 Z30'0 8t'10 SO0'0 °I. UP, h'ob 9'9, t'8 O'C '1 'F. OTLLO ta0'0 h00'0 6E1'0 '9 E'9F L'LE h'C1 1'6 O'h O'f 'F StLLO 't100'0

fl'O

770'0 CO0'0

CWO L70'0

"4 'I.

9'7,4 Vit, I'll 7'6 I'll

O'r O'S

O'f O'f

'0

'0

'1

'Z

htLLO fILLO

0'0 400'0 171'0 '9 S'Lh O'h O'f 'h ZILLO zovo 66L'0 100'0 'Z f ?t, O'rh L'L VA V( VZ '7 OILLO 4hO*0 180'0 SO0'0 '9 E'Zh Fhh fE 0 SOLLO b00*0 710'0 IWO 9 tIlf F'Oh UR 0'4 O'r - '1 'F h0/LO tOO'0 P00'0 CO1'0 'I. 7'6f 1.6 R7. 1 *b FGLLO ZOO*0 '150'0 Ot0*0 ft 0*Ott t*hh 1'71 0*L 0 *7 IOLLO LIO*0 66r0 0E0'0 'Z 6'81 7'ff 6't1 1'71 C'F C*7 '0 *7 469LO Ot0'0 060'0 ZWO 'Z l'6E 4'8E 6'6 7'6 6'Z 7'7 't 669/0 Z30'0 S44'0 Z90'0 .1 8'8E 9E C6 .4'6 0'7 O'Z 'C "410

(ET./ (ZT) (TT) (0T) (6) (8) (L) (9) (S) (7) (E) (z) (I)

(PanuT4u0D) aTiVI

17ZE

325

Construction of ,the Industry Data Base

Data on industry, characterigtics were gathered from numerous

sources including the Internal Revenue Service's Corporation Source

Book of Statistics et Income; the Department of Labor's Employment and

Earnings for the United States 1909-1968; the United States Census

ume on "Industry Characteristics"; and the Department of Conmerce's

Ccrsus of Manufacturers a;4 Mineral Industries and the Input-Output

MLtrix for 1958.7

.::ach of these sources were fully investigated and dozens of

industry characteristics were computed before choosing the final set

characteristics included in the regression analysis. AG.in there

was a problem of matching identification codes for at least four

different ceding systems are used in compiling industry informatioh:

the Standard Industrial Classification (SIC), the Standard Enterprise

Classification (SEC), the SEC survey codes, and. the 81 industry

breakdowp usedf_.1 the 1958 Input-Output matrix. It was necessary to

rely 'on industry titleg to merge the several sources of data, again

checking with manpower experts to see that our conversion routine was

7U.S. Department of the Treasury, Internal Revenue Service,

Corporation Source Book of Statistics of Income: 1953 1958, 1961, 1965

(Washington, D.C.: Government Printing Office); U.S. Departmenttf

Labor, Bureau of Labor Statistics, Employment and Earnings for the

United States 1909-1968:(Washington, D.C.: Government Printing Office,

19.68); U.S. Census Bureau, United States Census of Population: 1960,"Industry Characteristics," Series PC(2) 7, 19660J.S. Department of

COmmerce, Census of Manufacturers and Mineral Industries:"1951, 1958,

.1963 (Washington, D.C.: :.7.-Ivernment Printing Office); and Wassily, W.

Leontief, "The Structure of the U.S. Economy," Scientific American,

April 1965.

Ob.

343

I

326'

proper. was made all the more difficult because of changes in

classification Over time even within the same data source. In a.9

good number of cases, for instance; the Statistics of Income had

different codes for apparently the same industry in 1965, 1958, and

1953. Data was collected for several points in time, namely 1953,

1958, 1961, and 1965 in order to obtain "historical" information on

such factors as concentration and after-tax profits. In theory wages

should be related to long -un industry performance, not fluctuations

in such variables. The years chosen reflect different poirts,in the

business cycle in the post-war period with 1958 and 1961 being

recession years and 1953 and 1965 being yeats,of relatively low

unemplOyment.

With the exception of the Source Book of Statistics of Income,

all of these sources are relatively well known to research economists.8

The Corporation Source Book is a set of unpublished'worksheet tables that form the basis for the annuallypublished iepotts, Statistics of Tncome, Corporation Income-Tax-Returns- Beginning with 1942, the Source Book providesdetailed industry group financial statistics by asset sizeclass. No Source Book-was produced for 1952, and for 1962,distribution by asset size Class was provided only for majorindustry groups.

' Assets, liabilities, income, deductions,tax liabilityand distributions to stockholders are provided for about 235industry groups. (These are broken down for each industry)Into 13 to 16 columns providing a total column and sizeclasses ranging from zero assets to assets of $250,000,000 ormore.

8Corporation Source Book of Statistics of Income, "General

Description," op. cit., pp. 1-2.

344

)

327

The Source Book is published annually for returns withaccounting periods ended between July 1 and June 30 of thefolldwing year. Therefore, the 1965 Source Book, for example,includes returns with accounting periods ended betweenJuly 1, 1965 and Salte-lo, abOut .one-half of which arefor the calendar year ending December 31, 1965.

The financial information has increased over the years,but since 1948 relatively few changes have occurred. Thenumber of asset size classes has also changed with datacurrently di-4:ributed by 14 size classes.

Of specite mcern to the researcher is the industry classifi7

cation system usedin the Statistics of Income. It has a number of

shortcomings.9

(

The industrial classification used was revised for 1963to conform with the Standard Enterprise Classification issuedby the Bureau of the midget. The structure qt the StandardEnterprise Classification follows closely along the linesof the Standard Industrial Classification, which was designed asa means of classifying separate establishments rather thanthe companies of which establishments were part.

Year to year comparability of Source Book statistics isaffected by consolidations and mergers* as well as'by changesin the law, the tax forms, awl-the industrial classificationsystems used over the years.

For corporations, industry statistics are greatlyaffected by the returns of the larger companies. Since areturn is classified by industry based on the activityaccounting for the largest percentage of total -:eceipts, thismeans that large corporations with\diversified activities areincluded in only one industry even 'though many of theirbusiness operations are unrelated to\the industry in whichthey are classified. It should therefore be noted in usingthe Source Book, especially for recent \,'ears, that statisticsfor an industry may be either understated by amounts reportedby corporations whose principal activity lies elsewhere, or

overstated by amounts reported by corporations classified inthe industry but having substantial operations in otherindustries.

9Ibid., pp. 2-3.

4

345

328

Unfortunately 'there is nothing that can be done about this

reporting problem and the exact biases are unknown. Attributing

part of the General Motors Corporation's profits from aircraft manu-

facture to the "automobile" industry may tend to either understate

or overstate the profit rate from auto production. For the sake of

present research we had to make the clearly inaccurate assumption

that the profit rate shown by General Motors came from motor vehicle

production alone while the reported profit rate in'the aircraft

industry came only from such primary suppliers as Boeing, Lockheed,

and McDonald-Douglas.

.The industry characteristics finally chosen included the

following:

Market Power Factor ("Concentration") - Most previous research

has used the four-firm concentration ratio to measure the extent of

oligopoly in a given industry. The major problem with this variable

is that it is only measured for manufecturing'industries, leaving

many sectors of economy without an indek of concentration. This is

particularly serious because the real variance in concentration

may be found between sectors of the economy rather than within the

manufacturing sector. We therefore turned to the Source Book to provide

an alternative measure of "market power" which would yield an index

for all sectors of the economy including retail trade, wholesale

trade, and services.

The measure we chose to calculate makes use of the asset size

classification in the Source Book. The "market power factor" (MPF)

346

329

is the percentage of total sales in an industry (i.e. "business

receipts") received hy those firms in the top two asset categories.

jnlike the traditional measure of concentration, the market power

'factor has a variable number of firms. This is apparently a minor

shortcoming for the simple correlation between four-firm concentration

ratios and the market power factor for all manufacturing industries

in 1965 is .89. What is lost in terms of a defined number of firm"s0

is more than gained by the ability to measure "concentration" in the

non-manufacturing sector.

The market power factor was calculated for 119 industries in

the Source Book for eactof three years: 1958, 1961, and 1965. (The

1953 data were not arranged in a manner which allowed calculation of

this measure.) Several weights were attached to the three years and

the several versions of the MPF were tested with.macro data to see

which explained average hourly earnings best. It turned out that

equal weights on each of tht years predicted wages as well as any

other MPF index. This "unweighted" market power factor was then

merged onto the SEO tape.

After-Tax Profit Rate - The profit measure also comes from the

Source Book The index of profits used is computed by dividing "net,

income" into "total assets." Thus the profit rate used is on an

assets b s rather than on sales. This measure is also historical,

using unweighted rates from 1953, 1958, 1961, and 1965. Again the

use of the unweighted measure came from ,,sts on macro industry data.

347

330

Capital/Labor Ratio - The actual measure used in "depreciable

assets" divided by the number of production workers in an industry.

Thil, was a more difficult and somewhat more unreliable measure than

the others for it relied on the merging of Source Book data on

assets with employment figures from Employment and Earnings. Insofar

as the SIC and SEC classifications do not link perfectly, the capital/

labor ratios are imperfectly measured. The data refer only to.the0

year 1965 because of the extremely time consuming process of matching

other years for the two sources.

Government Demand - This was measured as the percentage of total

final demand in each industry purchased by the federal, state, and

local governments combined. It is derived from the 1958 Input-Output

Matrix in a straight-forward manner. Its main drawback is that the

Input-Output industry breakdown is quite different from the SIC and

SEC classifications which poses soma difficult linking problems.

Percent Minority Employment - The final industry variables

used in the regression analysis were computed from the 1960 Censds

volume on "Industry Characteristics." The percentage of black male,

black female, and white female employment in each census industry was

calculated and added to the SE0 file.

All of the industry data is reported in Table A.8.

348

.01016

01017

01011

.02126

02136

02146

02156

03000

04706

04207_sAwmiits.PLANING

04208

04209

TABLE A.8

TBE INDUSTRY

___INDUSTRY

CONC.

AGRICULTURE

0.1021

DATA

PROF.

0.0164

0.0240

0.0059

0.0344.

0.0134

0.0414

0..0243

0.0245

0.0406

0.0406

0.0360

0.0419

K/L

GO

V. D

.

____NA_:_0.0240 0.2067.

NA

-0.1209 0.2155

NA

-0.12°4 0.1165

39.2072_0.1091 3.0420_

19.7765

0.0221 0.0560

23.5(29

0.0

0.1149

25.083170.0010 0.1076

11.0

209.

NA

0.1265

NA

0.0009 0.2747

12.8503

_0.0

009

0.2288.

6.1935

0.0009 0.3172

3.9548

0.0495 0.2439

FORESTRY

FISHEPTES

METAL MINING

COAL. MINING

PETROLEUM C GAS EXTRACTION

NlvmETALIC MININGF._OUARRYING

0.2484

0.0923

_____0.6063

0.3224

0.4425

0.1488

CONSTb,OCTION

tmlnimc,

MILLS__

MISC wnno PPooucis

FURNITURE E FIXTUPES

0.0216

0.2171

0.2371

0.1775

0.0693

04216 GI ASS F GLASS PRODUCTS

0.5577

0.0750

19.3688_0.0011

0.2855_

04217

CFMENT.CONCRETE,GYPSUM,PLASTER

0.2525

0.0528

29.8127

0.0010

0.1583

04218

STRUCTURAL CLAY PRODUCTS

0.5234

0.01°0

13.0643

0.0010

0.2626

_04219

POTTERY E PRATED PRODUCTS

0.2387

Q.0262

4.5705

0.3622

04?36

MISC MINERAL & STnNF PRODUCTS

0.4633

0.0544

_0.0010

18.1580

0.0010

0.2279

04737

BLAST FURNACFS.STEEL WORKSI1)

0.7180

0.0481

31.8519

0.0061

0.1639

04239

PRIMARY NONFERROUS INDUSTRIES,

0.5141

0.0316

69.3857

0.0525

0.2006

04246

CHTLERY.HAND TnOLS,HARDWARE

0.1581

0.0754

6.0327

MA

0.3280

04247

FABRICATED STRUCTUhAL METAL PRODUCTS 0.1044

0.020?

6.7429

0.0

0.1402

_0424,1MISC.FABRICATEO METAL PRODUCTS._____ _0.2372

0.0512

11.1692_0.0257

0.2370

04256

FARM MACHINERY C EQUIP

0.7339

0.0348

17.2499

0.00970.476

04257

OFFICF,COMPUTING.ACCTING MACHINERY

0.7267

0.0741

34.6225

0.0737

0.2534

04258

MISC MACHINERY

0.0545

9.3115_0.0502

0.1520

04259

ELECTRICAL MACHEOUIP.SUPPLIES

_0.1816

0.5204

0.0553

7.8971

0.0973

0.3735

04267

mnTnR vEHICALS.mOT VEH EQUIP

0.8786

0.770

23.4105

0.0332

0.2053

04261

AIRCRAFT C PARTS

0.8406

0.0426

11.4435

0.5137

0.1931

04269

SHIP C 9OAT BUILDING F. REPAIRING

0.3928

0.0149

3.8691

0.1865

0.1546

04276

RAILROAD F.

MISC TRANSPORTATION EQUIP 0.4538

0.0289

14.4807

0.1865

0.1627

_042862ROFESSIONAL

EQUIP C SUPPLIES

0.2547

0.0557

9.3756

0.0972

0.3210

04287

PHOTOGPAPIIIC EQUIP

F. SUPPLIES

0.6679

0.0107

37,1984

0.0972

0.2891

04287

WATCHES E CLOCKSI2)

0.61)5

0.0339

4.0E

15-2

--NA

.0.5313

.05306

MEAT PRODUCTS

0.3402

0.0235

8.8045

0.0072

0.3639

05307

DAIRY PRODUCTS

0.4055

0.0481

21.7709

0.0072

0.1635

05308 CANNING C PRFSPPVING-FRUIT C VEG

0.2748

0.0427

9.2858

0.0072

0.4929

.05309_GRAIN-m111

PRODUCTS

.0.3866.0.0525

27.9480_0.0072

0.2288.

05316 BAKERY PRODUCTS

0.2506

0.0538

11.0182

0.0072

0.3278

05317 CONFECTIONARY C RELATED

PRODUCTS

0.2460

0.0765

5.7786

0.0072

0.5652

115.

31.1

.14E

VE

RA

5LIN

DU

STR

IES

0.3492

0.0472

__19

16ji

IA...

.0.0

072_

0.

1866

.

cir

Ino

TABLE A.8 (Continued)

INDUSTRY

05319 MISC FOOD PREPARATIONS

o5,24-rnou:cn mANUFACTORES

05346 KNTTTING

(15147 DYEING C'FINIcHING TEXTILES(3)

05348 FLOOR CriVERINGIEXC HARD SURFACE)

05349 YARI.THREAD C FABRIC MILLS

05356 MISC TEXTILE MILL PRMUCTS

_

05 366 t""ARFL E ACCESSDRIES

05367 MISC FABRICATED TEXTILE PRODUCTS

05386 DIJt.Q.PA ()FR

C PAPERBOARD MILLS

05387 PAPFPBDARD CONTAINERS 1

BOXES

05389 MISC PAPER C PULP PRODUCTS

05196 NEWSPAPER PuPLIcHING C PRINTING

05398 PFINTING.PUBLISHING.ETC(4)

05406 SYNTHPTIC FIBERS

05407 DPuGS E mFDICINES

05408 PAINTS.VARNISHES C RELATED PRODUCTS

05409 MISC CHEMICALS E ALLIED PRODUCTS

05416 PFT:4111FuM REFINING

05419 VT SC PETPDLEUm E COALORODUCTS

05476 RUBBER PPOOUCTS

05429 MISC PLASTIC PRODUCTS

05436 LFATHER-TAINED.FINISHED.CURRYED

05437 FOOTWFAR.EXCEPT RUBBER

'

_05438 LEATHER PRonucTs.Fxc FOOTWEAl?

0(2,506 RR C RA1LwAY EXPRESS SERVICE

06507 STREET RAILWAYS $ BUS LINES

06508 TAXICAB SrPV10E

06509 TRCCKING SERVICE

06516 wARPHDUSING r, SMRAGE

06517 WATER TRANSPORTATION

06518 AIR TRANSPORTATION

06519 PETROIrUu E GASOLINE PIPELINE

_06526 SFRvICES INCIDENTAL' TO TRANSpORIL.

07536 RADIO PPDAOCASTING C T.V.

01538 TELFPI-ONF(w1RF C RADIO)

_07511_TFLEGRAPHIwIRE E ROM)

08567 ELECTRIC LIGPT C PEIWFR

08568 GAS

C.STFAM SUPPLY SYSTEMS

.08569.FLECTRIC-GAS UTILITIES.

---.

CONC.

PROF..

Grytii. THIN_

------

0.3525 0:0467 _25.0241_

0.8967 0.0138

12.2338

0.1323 0.0382

4.0431

0.1744 0.0234_ 3.6054_

0.3168 0.0241

10.5136

0.3871 0:0345

33.4269

0.3947 0.0301.

2.4748

0.0449 0.0283

1.1906

0.1043 0.0247

2.2997

0.5858 0.0504

17.4776

0.1559 0.0556

8.1469

0.2147 0.0638

12.1658

.0.2117 0.0666

14.8525

0.0929 0.0478

9.3934

0.7331 0.0881

52.6891

0.6239 0.0993

32.4553

0.3819 0.0624

25.3242

0.5747 0.0572. 50.7991

0.9347 0,0320.510.8074

0.3071 0.0422

26.8645

0.6173 0.0522

17.1447

0.0072_0.3326

0.0

0.6173

0.0024 0.6976

0.0024 01.2761.

0.0024 '0.3845

0.0062 0.4470

0.0458 0.3890

0.0091 0.7753

0.0'58 0,6072

0.0089 0.1839

0.0010 0.3359

0.0089 0.3878

0.0213 0.2056

0.0213 0.369

0.0011 0.2633

q.0471 0.3142

0.0012 0.2364

0.0826 0.2287

0.0641 0.1497

0.0641 0.2750

0.0282 0.3063

.0.0212 0.3937_

0.0

0.2174

0.0080 0.5543

0.0080 0.5577.

NA

0.1370

NA

0.2139

NA _0.2163

NA

0.1246

NA

0.2969

NA

0.1980

NA

0.2556

NA

0.3/91

NA

0.3122

0.0

0.2709

0.0402 0.5737

0.0402_9.4325_

0.0415 0.1687

0.0415 0.1896

_9.0415 0.1687

0.0960 0.0407

5.3437,

0.0873 0.0281

9.4909

0.2468 0.0375

2.1882

0.0179 0.0317 _3.5414_

0.9257 0.0095

NA

0.1476 0.0140

22.7297

0.2833 0.0429

NA

0.0275 0.0384

8.1359

0.9275 0.0384

91.1336

0.2178 0.0160

NA

0.7105 0.0191

NA

0.3750 0.0373 217.5037

.0.1613 0.0216._

NA

0.5348 0.0622

16.9890

0.9590 0.0431

75.9472

0.7835 0.0264

38.3041

0.9442 0.0200 162.8866

0.8295 0.0229 169.9157

0040T 0.0268 290.7214

10

TABLE A.8 (Continued)

_INDUSTRY..

- _

08576 WATER SUPPLY

08578 SANITARY SERVICES

C9606 MOTOR VEHICLES C EQUIPMENT

_09607..PRUGS CHFmICALS E ALLIED PRODUCTS_

CoNC._PRoF2___K/L

0.1402 0.0194_71.5791

0.0415 0.2031

0.1402 0.0174

71.5791

0.04150.2980

0.3350 0.0433

2.6735

0.0090 0.190

0.2774 0.0565_2.7972

0.0000 0.3019

01608 DPY nonOs C APPAREL

0.0349

0.0233

2.0838

0.0090

0.3559

09609 Fnon F. RTiATED PRODUCTS

0.0146

0.0293

3.7193

0.0090

0.2879

09616 rAP4 PRODUCTS -RAW MATERIALS

0.2198

0.07,07.

NA

0.0090

0.2213

09617 riFCTPTCAL OrnnS n HARDWARE1-51

0.0903

0.3318-2.8999

0.00y0

0.2440

00618 mtCHINFRY.TQUIPFENT C SUPPLIES

0.0590

0.0391

2.4919

NA

0.2110

_10636 Font) STORFS1ExC DAIRY pRonuCTst_.

0.0496

3.4499

0.3685.

10637 DAIRY PRODUCTS C MILK RETAILINC

0.3882

0.0496

__.0.0090

NA

0.0090

0.3685

10638 GENERAL MERCHANDISE RETAILING

0.4285

0.0378

4.1517

0.0 '390

0.6920

_10619 LIMITED PRICE VARIETY STORES

0.6344

0.0436

0.0 90

0.8130

10646 APPAREL

E ACCESSUR IFS STORES

0.0668

0.0274_3.9344

2.4320

0.0u>)

0.e919

10048 EuRNITUIE C FURNISHINGS STORES

0.0632

0.0126

3.5125

0.0090

0-0,019

_J06,. HnusFunin ApPLIAmCES,T.v. E RA0.10

0.0632

0.0126

NA

0.0090

0.2717

10656 mOT08,VFHICLE C ACCESSORIES

0.0198

0.0133

2.8805

0.0090

0.1778

10657 GAS0LINE SFRVICE STATIONS

0.0195

0.0328

NA

0.0090

0.1096

_10E58 nPun STiRgs

10(5; EATING C DRINKING-

_ _ _

PLACES

0.1505

0.0573

0.0482_1.9740_ ...0.0090

0.0194

1.6875

0.0090

0.5399

0.6358

10666 HA8D,AARE E FARM EQUIP STORES

0.0375

0.0130

0.8246

0.0090

0.2182

.10676_044FR): BuILDINA.mATERIALS_

Cs.0427

0.0254

0.0090

0.2043

10678 LIQUOR STORES

0.1220

0.0260

_2.5258.

NA

0.0090

0.2622

10686fiFWELRY STORES

0.0518

0.0184

NA

0.0090

0.4470

_11706 BANKING

F.CREDIT. AGENCIES

0.5180

0.0046

1.0236_ 2,00710.58-44

6.3189--

11716 SHUR r. CIMMOD BROKERAGE CO (6)

0.0649

0.0136

1.5902

NA

11726 INSURANCE

0.7592

0.0115

NA

0.0071

0.4651

_11136 REAL ESTATE (7)

0,.0332

0.0127

NA

0.3957

12806 ADVERTISING

6.1818

0.0531

_0.0061

NA

(1.'04620.3954

12807 MISC BUSINESS SERVICES

0.0560

0.0377

NA

0.0462

0.4105

1780.11_NYo ,r,Fr!AtR SERVICES E GARAGES

0.0865

0.0232

NA

0.0762.0.1606

12809 MISC REPAIR SERVICES

0.0620

0.0398

NA

0.0271

0.1243.

13926 HOTEL

E LODGING PLACES

0.1337

0.0031

11.4702

0.0271

0.6004

_13826_LAUNDERTNG,CLEANING C DYEING

0.0747

0.0338

3.0no

0.0271

0.6433.

13836 SHOE REPAIR SHOPS

0.0620

0.0398

NA

0.0271

0.2763

13838 RARnER

F.BEAUTY SHOPS

0.0355

0.0505

NA

0.0271

0.6110

_14846 THEATERS

F. MOTION PICTURES

0.1062

0.0132

NA___ 0.0

0.3846

14848 pnwuNC AiLf7,6000l C BILLIARDS

0.0473

0.0169

NA

0.0

1481,9 MISC ENTERTAINMENT F RECREATION

'0.0473

0.0169

NA

-0.0050

0.4045

APPENDIX B

MEANS, STANDARD DEVIATIONS, AND CORRELATION MATRICES

352

VARIADLE

SCHCOL 1

WAGE 2

UN MEM 3

MIG<50 4

S.AT16 S

EXP' 6

TRAIN 7

PHY DM 8

NEG WT 9

GED- 10

.3VP 11

UMOCC 12

%iFCCC 13

.caocc 14

DA/PW 15

NYL/PW 16

AVEMPF 17

FUSL 16

ITSLEX 19

ATPEAS 20

MININD 21

EDS16 22

UNXMPF 23

%FMOCC 24MINOCC 25

335

Occupation Stratum 1-3 White Males

MEAL!

9.0507246382.707318841

0.57971014490.47101449280.253623188429.63768116

0.9420289855D-013.2971014491.3115942031.9521739132.292028986

0.15378985510.14364492750.3139130435D-01

STANDARD DEVIATION

2.8369883890.87618411210.49540360610.50097756970.436669448612.00604594

.0.29317494740.67714967550.76225486320.15341246880.5531675997048134619654D-010.19423951660.6059900926D-01

24.527083603.032517139

49.003754873.687716972

0.3629292271 44

0.4064332987D -01 0.6818532,97D-010.66475136461 -01 0.8142093 52D-010.41674e3766D-01 0.1734286728D-010.3012124489 0.16325534051.956521739 3.7569780630.461270531 0.30737459640.1750362319 0.23606847980.3288260870 0.2045273673

35a

CORRELATION MATRIX

VAR/ABLE

RUBBER

12

34

56

7

11.00000

0.248569

0.568278D-01

-0.109377

-0.275604

-0.476702

0.310148

-0.102892

21.00000

0.282418

-0.141940

-0.350388

-0.231745

0.189954

0.6888600-02

31.00000

- 0.788542D-01 -0.212231

0.6118280-02

-0.7720580-01

0.703220D-01

41.00000

-0.4956590-01

0.100180

-0.155215

0.100879

51.00000

-0.4620760-02

-0.1693970-01

0.1484710-01

61.00000

-0.355209

0.3219180-01

71.00000

-0.105241

81.00000

VARIABLE

910

11

12

13

14

15

16

iUBEIER

1-0.951219D-01

0.7605320-01

0.8630670-01

-0.609657D-01

0.6071280-01

-0.4066340-01

-0.165087

-0.4096430-01

2-0.126938

-0.3984180-71

0.8906610-01

-0.9550010-01

-0.193301

- 0.272B72

0.1809680-01

0.137444

30.2073020-01

-0.2630710-01

-0.2030490-01

-0.8516380-01

0.3674450-01

-0.11653d

0.4993440-01

0.262830

40.5249540-01

-0.3716340-01

-0.5746930-01

0.103286

-0.7245440-01

-0.4795090-01

-0.2809490-01

-0.3903520-01

50.2399520-01

0.6253340-01

0.446924D-01

0.473355D-01

-0.3567710-01

-0.448784D-01

-0.3973280-01

-0.130759

6-0.1229910-01

-0.9080330-02

-0.8363710-01

0.475820D-01

-0.122128

0.4555370-01

0.8251410-01

0.878570D-01

7-0.132308

0.133360

0.171196

-0.677227D -01

0.2481740-01

-0.106858

-0.5981570-01

0.2770470-02

80.696109

-0.740524

-0.728280

0.824973

-0.558752

-0.478152

0.192873

0.10988t

1.00000

-0.701810

-0.546288

0.683003

-0.167309

0.107956

0.5127750-01

-0.6897480-01

10

1.00000

0.737764

-0.675669

0.29104/

0.1511508

-0.187357

-0.5251440-01

1I

1.00000

-0.713320

0.139367

0.104047

-0.231765

-0.9494020-01

12

1.00000

-0.570883

-4250883

0.180390

0.1955770-01

1314

1.0000

0.600590

T1.00000

-0.157098

-0.149111

-0.1n0514

-0.232338

15

1.00000

0.766573

16

1.00000

L.)

'1/4

...1

C.7%

VARIABLE

NUMBER

17

18

19

20

21

22

'

23

10.141655

-0.5223770-02

0.4286880 -C1

0.149627

-0.6506290-01

0.1253540-01

0.7897730-01

0.3951680-01

20.275904

0.8780920-01

0.127912

0.263330

-0.381537

-0.280848

0.319694

-0.229096

30.273440

0.101074

0.115482

0.323286

-0.159091

-0.161839

0.689850

0.3184020-03

4-0.1034050-01

-0.2695420-01

-0.3071950-01

-0.9623670-01

-0.2664350-01

-0.8599330-01

-0.7296330-01

-0.7192530-01

5-0.152285

0.190044

0.207108

-0.128781

0.1938000-01

0.896623

-0.226826

-0.4087580-01

6-0.3509860-01

-0.3793360-01

-0.112249

-0.1449090-01

0.6247850-01

-0.146316

0.3041000-01

-0.8879450-01

70.8894890-01

-0.7855891 -01

-0.4844580-01

0.3263060-01

-0.9423040-01

0.7001520-01

-0.1695480-01

-0.7010470-02

a0.145505

-0.109526

-0.107072

0.5882310-01

-0.404059

-0.5513790-01

0.7874310-01

-0.582489

9-0.6525230-01

-0.182528

-0.206977

-0.102871

-0.4154.690-01

-0.2072310-01

-0.2603830-01

-0.109951

10

-0.4711090-01

0.204042

0.229803

0.3227020-01

0.198341

0.133140

-0.1165660-01

0.280161

11

-0.2350870-01

0.263119

0.275054

0.3880520-01

0.8271320-01

0.8588190-01

0.1012360-01

0.141381,

12

-0.3354200-01

-0.188988

-0.241124

-0.163714

-0.184729

0.9833900-02

-0.4489130-01

-0.534130

13

-0.124535

-0.9581650-01

-0.5314170-01

-0.3785580-01

0.446750

0.3443690-01

-0.6286950-01

0.979036

14

-0.249226

-0.108167

-0.160696

-0.281522

0.68'370

-0.9767410-02-0.196532

0.757455

15

0.297717

0.8908200-02

-0.5022800-02

-0.104221

-0.173451

-0.6675610-01

0.277756

-0.167538

16

0.627292

0.9947420-01

0.146574

0.417685

-0.305693

-0.128465

0.605029

-0.142350

17

1.00000

0.142604

0.219425

0.487101

-0.370070

-0.110573

0.762784

-0.166445

18

1.00000

0.947289

-0.9668490-02

-0.230682

0.180700

0.162758

-0.106605

19

1.00000

0.116404

-0.282053

0.223627

0.201473

-0.8497650-01

20

1.00000

-0.375978

-0.102924

0.446164

-0.103415

.21

1.00000

0.5846090-01

-0.320200

0.524530

22

1.00000

-0.183742

0.2580300-01

23

1.00000

-0.102209

24

1.00000

VAMIAULE

23

SUMBER

10.213631D-01

2-0.302409

3-0.3350450-01

4-0.4193730-01

5-0.2835280-.01

6-0.8356320-01

7-0.3502680-01

a-0.344233

90.144742

10

0.546339D-01

11

-0.720523

12

-0.218773

13

0:902962

14

0.774483

15

-0.121629

16

-0.156524

17

-0.205454

18

-0.198211

19

-0.193983

20

-0.184477

21

0.531948

22

0.336934D-01

23

-0.135826

2*

0.941776

25

1.00000

0 0 rt

338

Occupation Stratum 1-3 Black Males

VARIABLE

scHrot.

m E A14.

7.632411067

STANDARD DEVIATION

3.405652531

WAGE 2 2.16'94C7115 0.8149864818

UNInN 3. `0.4743083004 0.500329?723

MIG<50 4 0.4584980237 0.4992622702

S. 16 5 0.7984189723 0.4019761088

EXP 6 33.10276680 12.64649523

TRAIN 7 0.94861660080-01. 0.2936045834

PHY OM 8 3:470355731 .0.6577609567.

MCG WT - 9 1.3813992C9 0.8493935201

SVP 10. 2.088932806 0.45.76033661

%OvilCr. 11 0.2539683794 0.1793019063

ZWFOCC 12 0.1003873518 0e1630923884

2:8FOCC 13 0.3325691'00D-01 0.64527350560-01

DA/PW 14 )6459372127. 33.26334179

AVFmPF IS 0.3519994730 0.2770429297

FDSL 16 0.32050215280 -01 0.73419058830-01

ATPrAS" 17 0.39'56719368D-01 0.19102449120-01

ZMIN1N IR 0.3175426489 0'.1715143466

EDSI6 19 5.806324111 4.232109816

UNXOPF 20 0.2322388669 0.3108318211

7,FMOCC 21 0.1336442688 0.2155'543424

ItMINOC 22 0.38761264E2 0.2119432580

356

C0R9EkAT1CJ mA73IX-

3,171341.E.

4000F9

I4,

2:6

li, 4" S 6

1

1.00000

41`,

2

0.176282

1.00000

C'

3 \

0;121358

0.496698

1.00000

4

-0.140870

-0.181349

-0.141208

1.00

000

5

-0.210970

-0.338/2.6

-0.252756

-0.111057

1.00000

6

-0.654943

-0.5499590102

-0.3595590-01

-0.9925220-01

0.132110

1.00000

7

0.265190

0.180873

0.4367000-01

-0.8132070-01

-0.5448790-02

-0.227068

8

- 0.159886

0.124134

0.103114

0.114056

0.4484690.01

,0.8719020-01

16

1.0000:

-0.8812070 -01

8

s 1.00000

9ii7I401.0

910

11

12

13

14

15

r6

603/R

1 2 1 4 5 6 7 8

10'

1112I3

14 15

16

- ('.7816410 -01

4.122912

-0.167566

0.150160

0.169774

-0.155645

0.3063910-01

0.9235640-02

4.7591211-11 -0.47994809 .1 -0.212127

-0.41r9170-01 -0.142056

0.155235

0.535418

0 0 39909

0.4661410-31 -1.3591130-01 -C.101843

=0.2(93740-01 -0.5914610-01 -0.116395

0.445357

40.3060210 -01

0.491010-01 -1.9060170-01

0.6443830-01 -0.8698791-01 -0.57e1170-01

-0.5775550-01 - 0.3268900 -01 -0.4518080-01

n C

lli(?4

..., .

0 7.

ti

0...1405150-01 -1.6179420-01

0.2606340-01

0.6069600-01

0.4071030-01

0002113050 -02 -0.156224.

0.2502940-01'

g

0.4075300-02 -0.:43852D-01 ,0.396385D-01 -2.4338730-01 -0.7577040-01

0.172505

0.3058360-01

0.3005450-01

- 0.143465

0.1270450-11

0.649r6,0-01

0.143465

0.173288

0.106997

--0.3424640-01

0.5419790-01 -0.531-7780-02

0.634806

-1.606213

0.4561740-01 -0.615905

- 0.629007

0.325216

0.5930070-01

'

..00000

-0.346672

0.37494-

-0.327993

.-:6323710-01

0.2310180-01

0 290238

.4 0.157582

0.530013D-01

(1) rt 1 a

1.90000

-0.198680

0.309674

1.00000

-0.454323

-0.270263

-0.211116

-0.261278D-01

.....

C.JO

-:;;;7g

,1.3

-0.109907

-0.102988

-0.9985640-01

i W1.00000

0.745956

- 0.186105

-0.197400

-0.540830-01

01.00000

-0.224251

-0.209997

-0.482009D-01

vAP148LE

17

140041-4

f -

C.10c797

18

1.131158

19

0.404111

20

0.9260329-01

210.393596

-0.324944

-0.191594

0.561715

30.407791

-1.205,162

.-0.119489

0.788142

40.5145800-01

-0.116212

-0.194444

-0.117332

..4

5-0.5459700-0i

0.3919220-01

C.6,.00739

-0.708562

6-0.5091501-01

-0.7317680-01

-0.334903

0.1199910-01

7C.129470

-1.1940460-01

0.168137

0.5534440-01

60.217897

-1.424305

f-0.6172990-01

0.229329

9C.109411

-1.361913

1

0.1571910-01

0.134127

10

-0.10507

0.241110'

0.1551070-01

-0.114688

11

I.C.136496

1.175399

-0.8990270-01

-0.120788

12

-0.9789410-01

0.474112

0.160317

-0.116263

13

-0.161396

.Y.602463

0.158492

-0.1:7699

14

.6196645

-0 527499

-0.9565470-01

0.5457960-01

IS

0.49558.5

- 0.411127

-0.119964

0.776040

16

-0.124086

-1.151605

0.2729900-01

0.P920720-03

r1.00000

-3.309779

0.1153210-01

0.528674

18

1.00000

0.128388

-0.354616

19

1.00000

-0.11721^3

1,.00000

21

-72

,

1.00000

1.00000

0.356132

-0.2552700-01

co

1.00000

0.8123480-01

w

21

22

-0.101881

-0.285108

4`Y1

7,:c

w ....

0.164443

0.2971480-01

-0.3890360-01 -0.132492

-0.831768.A-01 -0.2974140 -01

0.5811060-01

0.8115010-01

-0.5550990 -01 -0.2376790-01

0:124312

0.6077060-02

,

-0.654326

-0.626883

-0.4093070-01 -0.334909

0.323468

0.160698

e.0.43588^

0.402604

0.979924

0.6/0576

0.865760

0.622387

-0.207141

-0.304464

- 0.736169

-0.327319

-0.5590570-01 -0.141336

-0.13176s.

0.540646

0.168864

-0.156794.

1.00000

-0.248459

0.698243

0.9568410-01

-0.261143

0.648288

1.00

000

J

340

Occupatiiin Stratum 1-3 White Females

kt.411AnLF

SCHOOLwArrE 2

UN ',FM 3

1q1C,<50ssaT16 5

FXP 6rwmpire n4 R

mFAN

9.()307692311.739646154

0.32307692310,5076923077.415104615433.23076123

0.3076n23077D-012.723076923

N1t wT 9 1.200000000GFO In 1.901538462"VP 11 2.3/5384615T,rl'AnCr 1? 0./510769211D-01ZwrOCC 13 0.4314000000l'AFrICC 14 0.1057692308.rdi/PW IS 11.074616474YL/PW 16 1.950009438AvrmPF 17 0.2597128205FOSt. In 0.25216372331)-61Fnst.f-'x 19 0.40577638370-01ATPPAS -20 0:40669230770-01MTNINO 71 0.4629771389E0S1( 27 3,353846154ONV,TF ?3 0.1493641026D-017,Fm0CC- 74 0.54416923P1M1'19 0C 75 0.6292769231

358

STANDARD DEVIATION

2.5554429090.46221329660.4712911888'0.50383147370,.496623212710.6576.9855

0.1740560530.64970407460.66614562970.7683/648973D-0140.45264888800.62124407180-010.22463940310.1?'79L49827124528014582.1700780455

048960114730./3392743295D-01.0j:545927902-2D-010.16366896120-01/0.17656814344.431530383

/0.18279561031 0.26196156.500.2201675859

CnaPcLartry mArDix

Vey

lnA

kc1

23

45

67

8O

N0).8E9

0 0

11.00000

0.451280

0.277039

-0.6086570-01

-0.318027

-0.569956

0.6810400-01

0.193433

21.00000

0.390739

- 0.121867

-0.6040770-01

-0.250028

0.158314

0.381390

(11

31.00000

0.2227180-01

-0.248545

-0.105287

0.6749720-01

0.245723

41.00000

-0.106639

0.7386550-01

-0.2741470r02

-0.8885690-01

0S

1.00000

0.9378560-01

-0.150188

-0.2533050-01

61.00000

-0.8812790-01

"1.00000.

-0.157610

Lei

7--,

0.7653410-01

8. \

-\

1.00000

VA91%a1f

imunFR

9in

11

12

li

14

15

16

rt C a

C"

I-P.2202110-11 -1.141100

-1.1)1121

-0.3242040-01

0.7605911-01 -0.171614

-0.21976:

-0.1559790-01 -0.5700460-02

0.294128

CIT

7-

C.)

102

ii,%

-000

5186

P-3

P-1.1/004

-0.217544

1.1146540-01 -1.I177o7

-0.,011046

0.8151110-01 -0.3511780-01 -0.220865

0.149524

0.9483670-01

0.249074

rt

4-0.7448799-01

1.3725690-02 -0.7545790-01 -0.133562

-0.215965

0.134298

0.153374

0.248288

0.203656

5-0.1899230-31 -9.4157730-01 -0.9833640-01

0.4866530-01

0.121683

-0.150719

0.8809430-01 -0.2258460-01

.,,

-0.200776

9.3581260-01

0.4651970-01 -0.1326560-01

0.7679660-01

0.143123

-0.2722190-01 -0.108715

70.898650-11 .0.4314310-01

0.7320250-01 -0.3201610-02

0.4018330-01

0.1025820-02

0.217408

0.144243

ft

0.166071

-0.542203

-0.569470

0.663882

-0.529709

-0.523451

0.9872320 -01

0.118=

91.00030

-1.354113

-0.2589810-01

0.540518

-0.300892

0.284041

-0.9310510-01 -0.9702190-01

11

1.00000

0.421421

- 0.672165

0.190967

.-0.185719

11

1.00000

-0.368233

0.407443

-0.116259

0.109440

0.3375280-01 -0.4887940-01 -0.5939920-01

-0.78'51

17

-0.7719800 -03

-0.149887

-0.124749

13

1.00000

0.5533480-01

0.218605

1.00000

0.136653

14 35

1.00000

-0.284549

-0.369998

16

1.00000

0.823977

1.00000

.P I-

e Ca cr)

v4411911

17

tq

19

20

21

22

Avn$4f4

I6.271762

A.11n41,

1.191157

0.739204

-0.231611

0.2661850-01

,1w-550.66-6i

9.133107

0.407/19

0.120011

0.45

4115

-1.556114

A0.1''"14!:

0.1''"14!:

,-1.146810-11 -0.1169470-01

0.2351'71

-0.272806

-0.137886

4.4

iG47

83-1

1- 1.1P8157

- 0.22)072

-0.123324

00.2199580-01 -0.172692

50.161118

-1.162115

-0.112125

-0.3317381-01 -0.3839780-01

0.904828

6-0.2797440-61 -1.176954

-0.246495

-0.416369

0.110637

-0.7685400-01

70.613156n-1i -1.115354

-0.9367870-01 -0.259974)-01 -0.3113470-01 -0.135894

80.1579/7

1.1599760-01

0.214485

0.416494

-0.366346

0.4541890-01

9-n.)49534

-1.7571710-01 -0.7305300-01 - 0.7823260 -01

0.307002

-0.1905460-01

10

-0.1

9744

41.859/510-81 -0.5935490-02 -0.196272

11

*-0.342506

1.73735?

0.5723930-01 -0.73/671

0.154775

--'0.100737

12

-1.1694610-52 -1.4314463-01

0.1620050-03

0.177114

0.3318780-01

0.5871430-01

13

0.115229

-1.141559

-0.261581

-0.177093

0.4459800-01

0.108633

14

-g:4=

-1.4822920-01 -0.750746

-0.653086

0.485321

-0.203332

IS

-0.276182

-0.883140-0/

0.305163

-0.438635

0.9085420-01

16

0.545518

-0.104143

0.129237

0.562.02

-0.409843

-0.1656340-01

17

1.00110

-0.104613

0.183446

0.535479

-0.507536

14

'

1.e0000

0.857257

0.175314

-0.139882

0.249562

-0.122891

19

1.00000

0.397594

-0.266499

-0.3347750-01

70

-0.623418

0.124842

71

1.00000

1.00000

-0.154003

77

1.00000

23

1.00000

74

.

-0.100739

1.00000

23

0.328252

0.419151

0.75/618

0.8311980-01

-0.6407680-01

-0.156543

0.189009

0.191578

-0.150410

-0.19/066

-0.272358

0.1513550-0!

0.2171110-01

-0.243927

0.409904

0.262693

,-0.6504950-01

0.505696

-0.6771860-01

-0.1372810-01

0.358497

-0.427783

0.7450020-01

24

-0.252509

-0.354991

-0.135868

0.188518

0.2935700-01

0.134862

0.3449910-01

-0.703774

-0.115882

0.437976

0.109076

-0.737656

0.873096

0.535114

0.460118D-01

-0.123562

-0.314227

-0.343756

-0.377601

0.274714

-0.7374350-02

'

C n rt O to

C) ft C B

VAQ1AriLF

75

NO14":11

1-0.778941

7-0.415044

1-0

.14

0.18o617

So.4465160-01

56.151076

70.461.0-11

9P.14610211 -ni

Inn.

3446

416.7 5 p7740-01

12 11 14 is

-C.595515

0.51

7212

0.594400 .

0.i964690 -11

15

-0.991.4050-31

17-0

.147

4cS

1.1

-P.1

461;

"719

-n.4

0a76

779

-C.4

1341

121

0.315277

22

9.7793140-02

21-6

.115

492

24

6.981695

2S1.

nono

n

0 0 ft

C.)

343

Occupation Stratum 1-3 Black Females

VARTABLP MEAN STANDARD DEVIATION

sr7v101 1 1081791825 3.014919250wAnr 2 1.158467153 0.5010061795woinm 3 0.2919708029 0.4563374645mtr.(c0 4 0.4160583942 0.4947122718S. 16 0.7449755474 0.4377280428Fyn 6 31.27007299-- 1147343713PIAIN 7 0.7299270073D-01 0.2610791095PHY Dm 2.343065693 0.5063995778

WI 1.186861314 0.7096827707SVP 10 2.411571832 0.3976152614'Put-ICC 11 0.71240876910-01 0.46008235440-01IlvFfICC 12 0.4319255474 0.9846640815D -01---'PF'Cr. 11 0,7868248179 0.1963778957nA/PW 14 7.171518952 14.38949504AW7mPF 15 0.1904223844 0.1584202399FrIcA 16 0.21408472950-01 0.11479490700-01ATP 4S 17 0.24441609E4D-01 0.17422144640-01sr.mt fl 0.5064407771 0.1071694530Erc16 19 6.204379562 4.462488557HNVAPF 23 0.41227493920-01 0.1063381657Fmnr, 21 0.7183501650 0.1871126820TMTNoc 27 0.7895912409 -,0.1492582915

361

'"0.^11'

%Atc 1

no,c,,Sqn 1.00000

3 4 5 6 7 8

-n.1116544-01 -0.117777 -0.8414690-01 -0.549443 0.235240 0.117308 0.771441 -1.278646 -0.251251 - 0.2.53636 0.158824 0.177/17 1.00000 -0.5340141-01 -0.286423 0.15/2480-01 0.128389 -0.8663030-01

1.00000 -0.116736 0.118669 -0.6607130-01 0.159821 1.00030 0.7240220-01 -0.2864800-01 -0.9927200-01

1.00000 -0.210368 -0.4264020-01 1.00000 0.8727980-01

11

V1.111C T.o,Cco

10 11 12 13 14 15

1.00000

16

1 -,.c71,c40 -n1 1.41,1491-4i 1.7)42550-01 0.172445 -0.263789 -0.3580980-01 0.3745700-01 -0.3079100-02 1111n .^114n,11

-1 .:10 711-11 -1.071,1,1-11

0.5195470-01 -1.111471

1,151 100 0.11'994

-0.2/7409 -0.7055/6D-02

0.6100190-01 -0.3419510 -01

0.9611240-01 -0.3750350-01

0.3132390-01 0.1564720-01

4 .1^7'111-1..17 -1.1?1..011-01 0.1762?70-01 -3.2117490-01 -0.2199160-01 0.115052 -0.3168800-01 S 1.100/450 -n -1.9757291-11 -0.4564310-01 -0.2150500 -02 0.156696 0.6110000-01

cp.177317 -0.7120020-01 -0.5094520 -015

_."4c01n-11 6,4609341 -01 0.101000 -0.124465 0.0798660-01 0.177848 0.8149260-02 0.2889390-01 -".144704n_11 -1.410.860)-01 -0.1143Q41-01 0.188130 -0.148184 -0.2140130-01 0.145039 -0.139114

4 ,071,4..1n-11 1.111110 0.491172 -0.117243 -0.595490 0.105344 0.274466 -0.272956 1.0m111 1.114254 1.514401 - 1.153294 -0.7279170-01 -0.297558 -0.348193 0.170842

1.00011 0.403151 0.245157 -'0.527995 -0.268350 -0.260920 -0.304105 1.00000 -0.272425 -0.684200 =0.8620500-01 -0.8219180-01 -0.4087370-01

1.00000 -0.342582 -0.5525270-01 -0.3124340-01 -0.255699 11 1.00000 0.8914000-01 -0.7260130-01 0.266827 !4 1.00000 0.509272 0.168116 IS 1.00000 -0.198695

1.00000

v4"t 1117 17 11 19 20 21 22

c

7

g

1^ 11 1'

14 IC

17

;

71

r.11-776 -1,4764^91-01 0.427105 '0.464170)-01 -0.146103 -0.211470 ..114<10 -1.1non,7 -n.4',456'0-11 0.194787 -0.211576 -0.249223

-.14211,01-12 1.50c.0470-11 -4.274499 1.605/61 0.5416420-01 0.3285570-01

P.11'4411'1-11 .1.170554 -0.,757/4 0.8717121-01 -0.4.142950-01 -0.3840190-01 .......1c4,..0 1

. I r-,1 11 0.817471 -1.176424 0.163124 0.178346

-^.117011 1,441e401-01 -0.?1,1605 -1.3547443-01 0.3209270-01 0.7105670-01 ^.17/144 -".Iti4i41-01 n.l"1704 1.302421 -0.5651970-01 -0.7696930-01 1%141,11 ..0091067 -0.7)1)100-01 1.157332 - 0.69119) -0.720173 -.141177 1.20o.470 -0.4141710 -01 -0.044510)-01 -0.157058 -0.3834080-01

^41.,csn-11 1.77S034 _0,109514 -3.145143 ,0.404078 -0.382289 1,,:577Jo -1.q17.470-11_-n.61,9070-01 -3.4851701-01 -0.862490 -0.772986 x.114547 1.111116 n:4401,.140-11 ".8601431 -01 0.156696 0.124999 .....51710 1.15e071 1.117,?400-01 -1.4116221-01 0.869236 0.878479

.^.1'101%cn-11 -1.44411A 0,01454n-01 .1.2521161-01 0.644/770-01 0.5425800-01 roinc49 -1.11x.717 _0.47.1170-11 0.407043 -0.9263780-01 -0.141468

_m5,*-120-11 -1.14^018 -n.2445410-01 -0.107282 0.145440 0.169777 1,014,s0 -^.264541 - 0.173617 3.211506 -0.602181 -0.584510

1.91900 ' 0.7422770-01' 0.3483584-01 0.253749 0.289834 1.00000 -1.104363 0.101165 0.105464

1.00000 -0.3991580-01 -0.6501400-01

. . 1.00000 0.987758

1.00000

1 Y"..

Zs

345

Occupation Stratum 1-3 Cross Race-Sex

VARI ABLE

SCHOOL 1

WAGE 2

UNION 3

MIG<53 4S 16 5

EXPRACESEXTRAIN /

MEAN

8.8050541522.266895337

0.50180505420.48014440430.411.552346630.91335743

0.267148014401.3285198556Os 79422382670-01

STANDA2D DEVIATION

2.9863657110.89103776410.50090171400.50050985400.4930355431

11.696946300.44327113650.47052504260.2708862774

PHY D,/ 10 3.158844765 0.7542412214NEG WI. 11 1.357430722 0.731331.2472SVP 12 2.251985560 0.5039388620BMOC: 13 0.1444620939 0.93681986960-01

04FOC: 14 0.2284296029 0.2387936304ZBFOC: 15 0.65581227440-01 0.4130958844DA/PW 16 19.81431531 37.11690734AVEMPF 17 3.3177830325 0.2611143641FDSL 18 0.32736540600-01 0.599998435870-01.ATPFAS 19 0.39776173290-01 0.17246831550-61%MININI 20 3.3590887916 0.1878449060EDS16 21 3.122743682 4.294345252UN,-;MPF 22 0.2012516652 0.2839610)08ZFMOC: 23 3.2940108333 0.3066052611tMINOC_ 24 0.4384,729242 0.2526728368 1'

RA *&EX 25 0.9386281598D -01 0.2921656195

363

CORRELATION MATRIX

VARIABLE

Numsal

1 2 3 4 5 6 7

1

1.10300

2

0.251772

1.30000

3

0.123764

0.392999

1.00000

4

- 3.7774350 -01

-1.9129260-.01

- 3.1069550 -01

1.30330

5

- 0.341516

..-0.378463

...0.223098

-3.113598

1.30003

S

- 0.522938

- 0.231373

-3.3893210.-01

0.3993220 -31

0.8348670 -01

1.33)31

7

- 0.132948

.0.243278

0.1413840...01

- 0.8666553-02

0.357206

3.1775740-01

1.00000

8

...0.6712210-33

- 0.483153

- 0.213359

0.4721010 -32

0.8656630 -01

0.144754

0.2934980-01

1.13333

VARIABLE

9

N4Matil.

10

11

12

13

15

16

13.278979

.-0.535448031

- 0.7581470 -01

3.6793950.-01

-3.112533

0.4255730-.01

-0.5:140473...31

-3.122453

23.194320

0.265983

- 0.9353510 -01

- 3.2108650 -01

3.1366.7

..-0.379372

- 3.411184

3.132669

3-.3.1e50390.02

0.141043

0.2295910-01

.-3.110896

0.2414940-01

-3.116945

- 0.176126

0.6527950-.31

..-3.6849670.01

0'.7556903-01

0.3430490-01

3.6628050...7.01

3.3628240 -01

- 3.1423250 -01

-.3.67:6543-.31

-0.2328853 -01

5-.3.134990

0.1843230 -01

0.7291910 -01

....3.8205180.-01

0.101616

3.3132330...31

0.3994700-01

-3.4031230-31

6-0.323284

....0.6660800.-01

....3.1796800.01

..-3.3285540....01

- 0.3798750 -01

0.3495680 -01

0.115627

0.5373683.-01

7 80.3387780-.01

- 3.6331790 -01

0.130194

- 0.484483

0.184998

..0.7921580...01

- 3.156471

3.143165

0,224333

- 3.451426

- 3.5657160 -01

0.617296

0.152236

0.472324

.

-3.1812380-.01

- 0.195158

91.03000

....0.26:0420..01

- 0.5235800 -01

3.136857

....0.8803510...03

3.3727530 -02

-0.6159100...01

- 3.3222820 -01

10

1.33000

0.507577

....3.632832

'0.735298

....0.671355

- 0.591319

3.243653

11

1.00000

-3.386817

0.492518

...0.231153

0.7799410 -01

2.659E370...01

12

1.00300

- 0.511932

0.216788

0.111965

- 3.228693

13

1.30000

- 0.693)7d

- 3.374247

3.179973

14

1.03)03

0.447923

t.-0.21801

'15

1.00000

-0.188547

1E.

.,.

1.03000

Cr,

vAA1.481.9

17

18

19

20

21

22

23

24

MumaEa

13.137572

0.601.53-01

0.217391

-0.5983190-01

0.4933610-01

3.116148

3.1456020-31 -0.24)4430-31

20.399659

0.137833

3.318402

-3.509472

-0.289846

0.441/73

-3.447294

- 0.503227

33.320646

0.884910-01

0.335484

-0.234721

-0.146645

0.707455

-0.156347

-0.18)401

40.5751480-31 -0.3752710-01 -0.6896590-02 -3.7141950-01

-3.152261

0.1333900 -01 -0.3596870-01 - 0.3319390-01

5-0.142854

0.5568710-01 -0.155356

3.7721130-01

0.07696

-0.193)81

0.3869690-31

0.6487483 -01

6-3.7778200-01 -0.7943490-01 -0.117378

-0.117372

-0.535;950-01

3.6989390-31

3.7171600-31

7-3.7668880-01 -0.8205453 -01 -3.151674

D.125203

3.5370380-01

3.277735

0.1271510 -01

0.4217430-02

0.8828130-31

8- 3.243260

-0.135043

-0.103418

3.521601

0.6575573-01 - 3.271759

3.654993

0.627427

93.138626

-0.7240730-01

0.3507240-01 -3.5998590-01 -3.2386923-01

0.8341840-01 -0.1981560-31 -0.2437150-31

10

0.304624

-0.1425273 -01

0.193480

-3.572653

-0.2505630-01

0.250554

-3.741145

-0.626719

11

-0.2937910-01 -0.101208

-0.5527970-01 -0.1903890-01

0.4251070-01

0.3173510-31 -3./51260

-0.9376120-33

12

- 3.159227

0.18)564

-0.7748710-01

0.228782

-0.5108423-01 -0.113561

0.210141

0.8518950-31

13

3.7119320 -01 -0.7944850-01 -0.2423800-01 -3.339414

3.2331950-01

0.9303710-01 -3.677837

-0.451755

14

-3.233384

-0.143196

-0.9144440-01

3.535278

0.i773660 -01 -0.232315

3.944054

0.883591

15

-3.323755

-0.8593160-01 -0.370892

3.619209

0.5299880-01 -0.253133

0.717721

0.732163

16

0.315218

-0.214538

17

1.03000

0.101143

0.526827

-3.448590

-0.8992020-01

0.773,355

-0.231791

-0.337894

0.1155500-01 -0.3449350-01 *.3.255320

-0.7338210-01

0.281233

1.00000

-3.394771

-0.5727480-01

0.451785

0.9328950-01

0.126107

:30.M42:72

-3.210980

-0.203253

--3.264999

19

1.'50000

-0.1330860-01 -3.233932

C.:

18

tin

20

1.00000

0.8403410-01 -0.398333

0.645295

0.657193

Cri

21

1.33003

-0.130543

3.8785793-0/

0.115257

22

-3.255045

1.00333

^0.274973

23

1.00000

.0.962133

24

1.00003

VAR1.481E

25

taim8E.1

t-0.6200400-01

2-1.301137

3-0.5057720-01

4-0.3676300-01

50.133308

60.7342203-01

70.533067

80.460135

90.4283500-01

10

-0.314533

11

0.4591230-01

.4--.4

12

0.129154

13

-0.216303

14

0.285380

15

0.474231

16

-3.124497

17

-3.207867

18

-0.8813860-01

19

-3.238292

20

0.394222

21

3.9474440-o1

22

-0.135682

..:3

0.397174

24

0.401751

25

1.03000

I

Cn 0

W .4".

.tr)

...4

1U

) 7J C, A (6 (..n

(1) X

348

Occupation Stratum 5 White Males

nal:ABLE MEAN STANDARD DEVIATION

SCHOOL 1 9.671171171 2.730147298WAGE 2 2.666146649 0.8895596554UN MEM 3 0,5990990991 C.4906336086

4 0.5112612613 0.5004370408_M1G<50S.AT16 5 0.2545045045 0.4360736756EXP 6 29.76603604 12.09207279TRAIN 7 0.1328828629 0.3398310226PHI DM 8 2.110360360 0.3478159601NEG WT 9 0.31531531530-01 0.1749463546GED 10 2.532882683 0.1126325025SyP 11 2.929729730 0.1680291250%6KOCC 12 0.7407432432D-01 0.4341615898D -01%41COC 13 0.M9346847 0.1113156607-A4irocc 14 0.2549549550D-01 0.6611944693D-02DA/PW 15 30.70714391 73.18803550NYL/PW 16 4.329228404 5.144872989AVIAN.. 17 0.4467765015 0.2752137339kDSL 18 0.3552091090D-01 0.4074778927D-01PDLEX 19 0.7086173176D-01 0.58541519980-01ATPFAS 20 0.4969414414D-01 0. 16 19400692D-0 1MININD 21 0.2845968977 0.1440114972EDS16- 22 2.317567568 4.251204310UNXMPF 23 0.3005274024 0.3264232519%FMOCC 24 0.3054301802 0.1134623060MINUCC 25 0,3765045045 0.9497692474D-01

:.366

00000'1 .

MIS/Sr° 00000"i1.0-0SE995b'0 o-aLELLWo 00000't /10-079EMS'0 tO-CtC4r09'0 E6C06Z'O 00000'110-0E9SZOE'0- ZO-aLLLLO9'0- 9999/.1'0 eflfl09'0- 00000'1

10-aZSLfl92'0- 1.9-069ZSICO t9COSS.0 9Ltal'O- SSZEWO 00000'1

10-ata9t6"0- tO-G060627'0- OESOtf'0- SLCM1'0- ZE67,SF.0 20-CL,0Lf/S'0 00000'1

1.0-016119Z4'0- to-aeitaSE'o- EL116S'O- 116L017"0- 10-09flUZL'0- 6E069E'0- r0-0,57.)69'0- CCOCO't .

to-acinffuo- t0-092ZS0L*0- 9901/.1'0- nssta'o- feOrtS0 10-asin58170- 9SE'0 10-09LLP01,'0tO-assu.SEo- t0-c695c0Vo- t0-ORR699Z'O 66ELL1'0 oh141Z1"0- to-nrr*opt'o- tp-n*Do65vo- i0-c0too2r'n-

6SE601'0- 1.0-096LLW0- 1.0-a0OUSC'0- FC9161.0- no9cPL'0 i0-(165cEWn opftwo 1C-CtP6eLf'0

10-a8tArt6N) 107(710059L'0 10-4716069010 to-0416000- t0-000q49r'r, 10-4970Lbs'0 10-00ILLsc'41- fo-e06f,70b,0-

10-06Z799C0- 10-COthS6t'0- 10-0Uptt7'0 70-096EOSS'C 10-0909t0t*0- 10 -O 04(060'0 70-att4tS7'0- le-cf7ros'o-

tna81'0 10-0thL8LA'0 70$111.0 1.0-06(0077'1., 10-01671.07'0 t0-nrtr7wo 11171-1 '0- 1 0-0 tzt Pei A "o-91(6SUO 069661"0 1.0-7,9118ZOS'0- 7t6Sit'0 696ft91'0- 10-rrrtc016'n- fLrt,CL'f.- tc-rcfC01"0-

hattSt'0 ltchOl'0 Z(riSt.o- EtoLWO 977FP7-0- atttf0.t'0- 777f41'0- tf-Gef7;1P'0

91 St 171

00000'1.Z0-0tWEE6'0 00000'110-0StL999'0 19LCZE.0- 00000'110-0E95915'0- t0-00SZZ0E°0 10-0919LSV010-050179E1'0- 10-eZ6U9V0- 10-0az9L19'0to-anozwo 10-0911(S17.0- to-azoSorC"o

199Z01'0- L6ftSWO fzEbt141'0-165811'0- 91,Z0LZ'0 6E0/69'0-

9 L 9

£1 71 11 Ot

00000'I811ZW0- 00000'1t16rrl'o-0-0806SECo 00300%7607P1'0- 961S0t'C- Ofn.fivel9h011t'0- t0-36S0919'0- 70-41996S1t'0

S

00030'a(91)9C0

i2r1ZP9

71e%Irl'A

1

0

Se.

C

Oconotrt

1.'rSalwlior:ryA

40T4Y1=4m0:1

YnaLA0Li..4811ci1

1

2

11

3.1146010.3J35460...1314

10

0.1117590.1):46,0.141441

19

0.17'26028.2119010.125909

4 -0.6570530-31 -0.4341)30-01 -0.4152540-01i 3.11/7)00-81 -0.311720-01 -0.1/55100-01o -0.9084/90-31 -8.1144.5 -0.1076157 0.411/508-3 2 0.3774200-01 -0.0260070-038 -J.45441l8 -J1 c.76)1(5u-01 3.8372)10-01V -0.322/14.3-31 -C.24/2030-01 -0.8'267900-02

10 -3.9069610-01 -0.111358 -0.112297II -0.119561 -0.5317630-01 -0.126176

12 -3.0234790-81 -0.111680 -0.15470313 ,0.1737664 C.100294 0.14240814 -0.1623190-31 -0.21.4907D-01 -0.5174150-0111 0.3uio36 O. 117125 0.1115060

I. 0.570244 0.240794 0.2414671713

1.00000 0.2342571.00000

0.2000350.8)0208

1* 1.00000

1021222324

i0

10'1=3 0.230514

0.514272D-01-0.1104470-01-0.162510-0.2791490-010.2311920-01-0.610052D-01-0.272362-0.5098100-01-0.09/6130-010.112076 .

0.7167100-01-0.11.03300.2477930.3599710.2156180.3176121.00000

VAidArLE 254034:65

1 J.0749850-812 3.1181)43 3.4,7679J-014 0.1347100-82S 0.s/1901J-42o -1.1421047 0. 15413J3 3.5,36840-31

-8.5,25101J -0.2v3470-C111 -0.4321058-J112 -.0.23364413 2.911101514 J.4»73815 3.02.4158-J210 3.4314328-J11/. 3.1+001/

1) 0.4430153-3128 3.4:343660-0121 -0.1272300-0122 3.2527.6D-0123 0.590003D-8124 3.92339625 1.00000

11

-0.153175-0./54646-0.2201100.4116100-010.7219230-01.0.61237130-01

-0.2657510-01-0.4307160-010.3427710-010.3540170.701203D-010.3726)20-01

-0.2133440-01-0.6626080-01-0.202394

%:°0:.'i31441Z491

-0.252498-0.223267-0.361329

1.00030

22

0.8545230-01-0.7401100-01-0.9701420-01-0.1412130.934362

-0.8335160-010.7541200-01-0.7710780-01-0.3777520-01-0.5493650-010.2404230-01

-0.2584420-010.2052150-010.5261210-010.1363420.1234970.644569D-01

-0.1976670-010.1608100-010.3123930 -020.4996493...011.00300

4V

23

0.6311910-010.2264120.7493920.6436640-03-0.2645800-01-0.1953470-01-0.3747643 -010.416955D-01-0.9481524 -01-0.120049-0.1907540-010.2188420-010.3420850-010.7327390-610.2809880.4648960.7059760.2803600.257986.0.367930- 0.3216080.9074440-02.1,00000

24

0.1644710.1696830.260836D-010.6692770-02-0.109839D-02-0.1900400.175727

-0.153758-0.434643-0.160028-0.102475-0.5824120.9984520.3517280.8314613-020.5271930-010.1537550.9713220-.010.136774

- 0.2490869 -010.3104780 -01-0.175560-011.00000

O00CIDCr

0

cr)

rt-t

rtC

%.J1

ti

rt(3

u

0rr

351

Occupation Stratum 5 Black Males

VARIABLE

SCHUOL 1

WAGE 2

UN MEM 3

MIG<50 4

S.AT16 5

EXP -6TRAIN 7

PHY OM 8

NEG WT 9

GED 10

SVP 11

MEAN_

- 8.3393501812.385234657

0.5523465704( 0.400722021.7

0.718411552330.02527076"0.12274368232.19/i9458480.14440433210-012.5516245492.946931408

STl iDARo OEVIATIOV_

_3.485009868y.80137090364981522915

0.49093173610.450587784712.65896359

0.3287366838"0.42337893820.11951348030.12411671920.1306324041

8M0CC 12 0.897833935003 -01 0.62065621060-01--7,WFOCC 13 0.2651732852 0.1161655441%3F000 14 0.26772563180-01 0.63970870520-02DA/PW 15 22.20990909 39.13718419NYL/PW 16 3.600650095 3.657186912AVEMPF 17 0.4236900120 0.2953215854FDSL 18 0.38481035710-01 0.68271206250-01"

FDSLEX 19 0.61,083640790-01 0.78777844050-01ATPFAS 20 0.45847653430-01 0.18336128010-01MINING 21 0.3057716861 0.1555410360EDS16 22 - 5.501805054 4.538582225UNXMPF 23 0.2807524669 0.3346285613%FMCCC 24 0.2919458484 0.1191812678MINUCC 25 0.3817292419 0.91073721790-01

369

C044ELATICN MATRIX

vAKtAbLE

NUMBLK

12

34

67

8

q1.00000

0.264135

0.139996

-0.103064

-0.312712

-0.565397

0.232328

-0.106389

21.00000

0.570692

-0.174990

-0.276255

-0.5536530-01

0.162318

-0.3099750-01

'

3_

,..

--

1.00000-

-0.108306 --

-0.240785

-0.4588760-01

0.115497

-0.4856030-01

41.00000

-0.126835

0.380089D-01

-0.81.7211

DI.-0.4600220-01

51.00000

0.161324

-0.132722

',..0.6061357U-01,

61

1.00000

-0:262814

0.175520

.7

1.00000

-0.1635250-01

,

8t4

1.00000

.41

NI

4:)

VAR150LE

tiumaHk

I 2 3 4.--

5'

b 7 8 910

121314

15 16

9

-0.2050710-01

-4.806t420-81

-0.134457

0.8627480-01

0.8501250-02

0.2610110-01

-0.4527780 -01

-0.5583660-01

1.00000

.-- ------.

10

-0.104309

-0.207776

-0.257756

0.1603540-01

0.1730750-02

0.175116

- 3.102585

0.552443

- 0.5043820 -01

1.00000

'---

11

-0.4943520-01

-0.48t2590-02

-o.igesoup-ca

0.3336740-01

0.2219970-01

-0.6601153-01

0.7629920-G1

-0.316697

0.414881

-0.246062

1.00000

".0.275046"-*'-`

12

;0.215242

-0.188413

-0.191349

0.670704J-01

0.9653340-01

0.229475

-0.6706013-01

0.710469

0.757755001

0.679542

1.00000

13

0.233809

0.2104 f1

0.1884'O

-0.140802

0.9034390-02

-0.t44747

0.4915710-01

-3.299122

0.249672

-0.363919

0./49786 - --

- 0.679247

1.00000

14

-0.1619u30-01

0.9562380-02

-0.8188600-02

-0.7792590-02

0.158707

0.2425840-01

0.2538330-01

-0.184235

- 0.450638

-0.105630

0.402164

0.4116440-01

0.450006

1.00000

15

-0.111416D

0.I62213-01

-0.126132

,

0.168385] 01

0.112849

0.4015351 02

-0.2252161 01

-0.69606st DI

0.2757731)"1

-0.115032

0.358716(..

.

-0.503984f-01

0.255081t.02

0.15519 1.-01

1.00000'

.

16

0.100481

0.334962

0.149023

-0.2048750-01

-0.104782

-0.6757560-01

0.5438470-01

0.3707150-01

-0.2816950-01

-0.20042

0.453568001

-3.147826

0.133718

-0.5687620-01

0.645331

1.00000

°

VA

8111

LE11

N,U

N8E

R

0.1J58-)

20.

487.

3,11

30.118745

0.95757,0 -02

5-0.7316390-31

60.4270030-01

70.132335

0.1',1544

9C.6L.2210-31

10

-0 195163

11

-0.9506010-01

12

-0.113676

13

9.9856930-31

14

;A19750450-31

15

(0.343411

16

0.696181

17

L.00000

18

1920

212Z23

24

114

0.5037510 -01

3:1/1617

.J.1210,85

0.2144200 -31

-0.121166

-0.5066420-02

0.305422J-31

0.2466040-01

-0.1553083-01

-3.401357u-01

0.5629300-C

- 0.109699

0.75542507.01

-0,128244 it

-0.341090042

0.165074

0.226,17

'

1.00000

et

19

20

21

0.9079440-01

0.22055

0.1649990-01

0.260958

0 43,-(22

-0.316254

0.149'54

0.6n4.58

-0.230515

0.1,1'9100-01 -0.1622860-01

0.6240110-31

-0.151844*

-0.265481

-0.353-..U-01

-0.4533740.01 -0 446843-01,-0.6117920-01

0.7067403-01

0.6394320-01 -0.9553490-02

0.10332,0-01

0.399168301

0.448.460-01

-0.2'9187W-01 -0.1035210-01

0.9581340-01

-0.8136763-01 -0.306390

0.345780

0.65449.0-02 -0..1223/1

-0.1320370 -01

-0.144810

-0:160854

0.130327

0.105!5:

0.111593

- 0.5573510 -01

- 0.1352.

-0.15:,842

-0.7387710-03

-0.164124.-01.-0.2343630-01 -0.200744

0.209108

0.576451

-0.335925

0.285080

0.542072

70.359535

0.963522

0.6948573-01 -0.272305

1.00000

0.172672

-0.291672

1.00000

1.00000

".

22

0.311124

-0.104256

-0.132652

-0.181,011

0.160311

-,.197103

-0.1714790-01

0.1101520T01

0.26.104J-0.

-0.3%93360-01

0.1010020-02

-0,4505530-01

0.151304

0.113139

0.2277260 -01

-0.6282330-01

-0.1215400-01

-0.104007

-0.117966

-0.13

87

0.26 5 70-02

1.(10000

VARIALSLE

h0303t8

.

2 3 4 5 6 7 a 910

11 12

1314 15

16171419

21

2223

2425

25

0.1504

0.14,155

U,10',444

-0.134434

0.8845750-01

0.2625740-01

0.1876279-01

0.8970150-01

-0.401752

-C.8503050-02

0.3186110-01

-C.182007

0.844222

0.672264

-C.3650560-01

0.654210D-01

0.4,91310-01

0.1258053-01

0.2670100-01

0.880960D-02

C.1767330-01

C.170232

0.130691

0.858945

1.00000

23

C.990766001

0..2..7324

0.5;0193

0.205,,49'

0.756678

0.163141

-0.530,2220.'01 -0.117667

-0.199149

'0.1732440-C1

0.612111001 -0.139564

-- .129609

0.4927540-01

4994520-01.-0.301442

-0.101:41

-0.267542

-0.20184

-0.360381

-0.5639000-01

0.167562

:0.174218

.-0.659850

0.198396

'0.998851

-0.5174040-01

0.492295

-0.167.2280-01 -0.1653260-02

0.447120

,0,127281

,

0.725547

0.9122710 =01

0.227042

0.6674750 -01

0.300084

0.500346

-0.338649

-0.1368

1.00000"--(1-

0.9581620-01

0.100915

-0.5436440-01

0,15354e

0.190596

1.00000

354

Occupation Stratum 5 White Females

VARIABLE

SCHOOLWAGEUN'MEMMIG<50S.AT16EXPTRAIN

1

2

3

4

5

6

7

PHY DM 8

NEG WT ` 9

GEO 10

SVP 11

%8MOCC 12

%WFOCC 13

V3FOCC 14

OA/PW 15

NYL/P4 16

AVEMPF 17

FOSL 18

FUSLEX 19

ATPFAS 20-MINING 21

E0S16 22

UNXMPF 23

7,FMOCC 24MINCCC 2.5

MEAN

9.9451627122.0097288140.41355932200.50847457630.274576271229.93220339

0.14576271190-012.010169492o.a2.604'4C67802.9203389830.46284745760-01U.45688135590.28936440680-0115.414216C42:1541274230;39423073450.29577245110-010.75153344610-010.44581355S30-010.45075372582.742372881

0.18641340510.48586779660.5321525424

372

STANDA/R0 DEVIATION

2.2731291950.-42205766610.49330815360.50077761260.447058877110.35141780

0.26315279720.51826992260.00.18739415320.18329900210.27311597660-01f'.24214270860.10933619400-0119.785093172.721771670

J.z336196155,0.33049555440-010.:5433353090010.1697)918860-010,1129861367 \4.639530068

0,29186721780.24796618870.22 3212915

355

Occupation Stratum 5 White Females

.4 .4 .4 .4 .- .- - .4 44 .40 0 C) 0 CD 0 C) C) 0 0I I I

1 ' 1

I I I

C5 C) C) 0 0 C) CI C) 0,, o, -400 4, r. 0 44. 07 44 .4 C) 77 4.. 44. = ,r 0 r- A el 0u " -. 4. 4. eft .4 P . c , N7 I) ... .1, C. O. C. ID .... c. NI Cr 0 0

CO 0. , ., .4 .4 c) 1.4 0 4c. 'or .4 ..7. 4. 44 0 0 .7 -1 uN r. v) tO .73 0.09O,00000 . 5 5 TN ..r no a 0 - c . - 4 4 . o 3 ,.., .0 4: 0.... f... 4. 0 t 1 0 .0 .".". 0 -.J 13 0 r- C) .r. ,..) 0 0 ..., e4 .4 c. 044 -4 .4 0. 444.44.4 44 J? .... :V ...4... ..40,3 .... . 1.... r4 .0. . . . . 1. .., t0000000000000o .40 C3 0 0 0 CI 0 0II III C7 C) C) 0 0 0 CD 0 0 C) 0 C) CD4 9)

I I I-

.4 ..I N 4 ..I ... 4 .. .40.000 0 00 0'11:1 I - 4 41- . .).

C5 0 .".. 0 0 00 0P,,4 13 43 "si 4 '0 rm N 7 .0 4.1 0 1.. rg -. 4, N .. PO,4 rt., A ... 1) fs 0 0 0 4. "s.1 4.,.% .4 4. 33 4 7 J 0 4 0t. r ....NN..! r 0 V* r s ON 7 - tO .0 0 rl .4 ... 0 .4 0-7-7.0-.0.00 .4 N. N. NT 4. 4. .4 oft P./ - P./ P.1 A 07 A " 1 A A 0 0 .0 ,,J .0 4 tO 7 ... .r ,-, str v 0 NO- 2 . ^ 4 .4 1 4 1 4 - . . n J1 .4 CO .-. ^. ..r. 0 ..r cr - ,r eu o

...Ir.. -t 00000 900. . . . 0 0 .40 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

I I I 1 1

.4 .4 .4 .4 ..... ...000 0 0 0 0I I I I I I I0 0 0 CD C) C) C)

01 4.10e.a3 0.4 .1 A ^,- 07 CD 1 N1 '2 7. '4741 S 2Z t; MO ':7. 43." T. 8'

,45 .-.. 0 J1, r00 ..). -4 J'0 .... 41 r. P. .4 ,"4 (.9 4' .t. 044 c) 4 CD .4. CI .4 41 ,r 4) 4. ... eu .1 ,,,I .. X) Cr 43 C)

V. ... e. . ) 44. 3 PAv n. O. r4 A 0 0 ,4 ?. 4, .0 7 .0 .- 0.... .4 .-.. Nt .4 A.. NO 44. 4 Cr tin

O 0.00 w4 o0ool000000loae w40 0 0 0 0 0 0 0 C._ 0 0 0 0 0 0 0 0 0

1 1 I i I ; I I

.4 . -. ru . .4

I I0 0 0 00 0 01111 1

0 -7.N 40f.?0

N2 2 R un 40 .4 44 4, C)N. CP 44)0 .........,^1..... !..4%.11 1... .70%0

IA In A7. 4.0 0 ,.., .., p.. un t. .1r1 V. O AOPIP N P. 0 ... , .. 077 CO -7 en to 4 77 00..., .4 +J a. 4. 44 r .0 4 C) 0 1)- A .4 C)-4 e4 Cr N 0

441 .4 CP .7 .0 r- Pa un 0 41 Ma,. . 0 . 0.4 099000000000 "4

0 0 0 0I I

0 0 0 0 0 0 0 0 0 0 0 0I I I , I I

N .4 .014 ,...* N . 4 ...4 .4

I I0000III)000

4. t...c 7ii :.J. ii

.., 0 ,r (, ,D 0 ... r cD .4 C)4) .4 .0 ..4 4. f... 7 Cr CP CP CI

en CI a ,1CI 0 CD

4, O'S. C) . 4.4

4. 44 .. ). ..C, .0 ,.4 1: 7 ..T. 00 " ., .1. T. 1` I, '1 rT J0

c). Pa in.n C5 -7 r4 4.4 e.. 0 ..i.' ..., CD., ... a% CD ON 0 7 4 7 ' .., 0 0 4.0 0 0 0 0 0.400000 0 00300000.000

I I I I I I

0 orlf0j NC) 00000000I I I I I I II IC) 0 0 C) 0 C) C) 0 00 0 0 a% A ,3 N un 44 en an CP 0CP .00 ..5 0 .4 c. I.- 45. )44. 4) un 0

R1 un NJ 0) rm. ,. , . N .4. c) 4) Cr C)..4 +0 . 7 f+ 0P .4 -0 r. an NOpg u1 0 44 44 4. .0 44 ra 'V 4: J00. Pa 44 I) ^4 ") N! J 4..i V, C) .44 .4 . . . . 0 4 -

C 0 0 0 0 0 0 0 0 0 0 0I I I / I I

.. 4- .-C) 0 C)I I I

C5 0 C3.1 0 v. 0 IN, .,, .. .4 .. 4) 0770 - O0 ,N .7 ,^1 P... 0

N ^'0 0 AN .11 ..) 41 ..! NT - V 0-4 CJ -. N N 4 0 ..) ../. !.. 77 0P 0 ,p ,,, zo oaniOUN 0

, - - < si ." ON CO.4 9 99 904

O 0 0 0 0 0 0 0 0 0i I I I I

... 0CC CD.. ..., Cf.

4 8X CD 0 0 0 0 0 0 0 0 0X 4 e

LI kJ C) 0 0 C) 0 0 0 0

.4 w4 ...k Cc 4 ,Y .4 A ..cs 455kutat 4 0 .4 IV el 4, Vs A P. 0 .1 na .... r.. " 4 N.0 P- 35 7 0 4 NI "I 4 uN AX - 1" - XCt eX 0 a 9O 4 Z 4X-/ 5. >

373

',439

14.a

lEh6

46t.t

t....

5 7 a 9

10

it12

13 15 te 17

'

1 19

20

212223 24tr

.LE

17

0.2391:72

0.401646

:.1616/7

0.1199460-31

-0.d227tIo-J2

-0.4111140-01

0.126699

-0.136326

6.0

u.s970.-3

C.66,1210-31

-0.230499

6.191091

-0.277639

0.705330

0.72

1180

1.00000

25

Id

-3.2311493-01

0.160513

0.1023.,4

-3.120251

-J.5648,3,5'5-01

-0.122...00-01

-J.7:3240 -01

0.146412

0.0

-J.294203

--.104144

J.119912

-0.242919

-0.134130

-0.251553

0.500987J-01

0.9014500-01

1.COLUO

19

-0.1710630-01

(s.t69671

0.8819430-01

-0.151561

-0.113631

-C..7502140-01

-0.6902220-01

0.312674

0.0

-0.4)3721

-0.113359

0.164376

-0.331169

-0.22129/

-0.270926

0.4950110-01

0.206.181

-0.812021

1.00000

20

0.106163

0.406870

0.199369

0.216/41) -01

-0.112642

-3.9186290-01

-0.311325J-02

0.140014

0.0

-0.275640

-0.102277

0.65'040-01

-0.246665

-0.331743

0.105538

0.639349

0.549196

.-351074

0.506030

1.00000

-0.1311600-01

-0.331060

,

-0.146241

0.8101790-01

0.154020

0-8413800 -01

J.4356240-01

-0.219991

0.3

0.2/2286

0.8550540-01

-0.4580220-01

0.201660

J.189869

-0.862/890-01

-U.394302

-0.416014

-0.321063

- 8.401020

-0.658W37.

1.00000

22

0.161542

.0.103524

-0.161350

0.4486220-01

0.962397

-0.131664

0.8824440-01

-0.9192620-01

0.0

0.1069960-01

-0.6211200-01

-0.1123220-01

-0.1169590-01

-0.9540480-01

0.9284760-01

0.1114760-01

0.1226160 -02

-0.35,7790-01

-0.111934

-0.108613

0.151716

1.00000

23

0.35067?0-01

0.35°742

0.761075

-0.26.16040-01

-0.905530-61

0.1324330-01

0.8414110-01

0.3114110-01

0.0

0.1569400-02

0.1526320-01

-0.9C41360-01

0.3770370-01

-0.136292

'0.351011

0.39/502

0.569496

0.121411

0.135169

0.372361

-0.286496

-0.7570820-01

1.00000

24

J.341119

3.13499A

0.9740160-01

-J.7235950-02

-0.6490310-01

0.7694750-01

0.191012

0.606641

0.0

0.874561

0.6252560-01

0.811458

0.999287

0.548414

0.362,683

4.100226

0.6108990-01

J.243128

0.333149-

0.254543

0.21,1.155

0.1563150-01

0.26i9640-01

1.00000

,, C lc:

C, rs 0 D

itn re

.n W rt C a %..n

10.

1 s

3141

20.1 43541

3-3.9 665650-31

- 0.674°^i40-02

S-0.6603630-01

6-0

,724

4840

-01

10.1da!i4

8-0.513115

0.3

10

0.878,04

11

C.11C121

12,

-L.7oat:3

13

6.094adt

14

0.5691,8

15

0.376912

160.4256 510-01

G.:64519

td

-C.1418,0o

t-C

.145

11 /

2J

-C.2'0,64

0.225old

22

-0.1a4a49:i-01

23

0.1790620-31

24

0.997449

25

1.00030

'

=1-

rt 0 tD a W to 0 rr

357

Occupation Stratum 5 Black Females

VARIA6LE

SCHOOL 1

WAGE 2

UN MEM 3

MIG<50 4,

S.AT16 tEXP 6

MEAN

10.132911391.863670886

0.46d35443040.39240506330.639240506325.34177215

STANDARD DEVIATION

2.5012211800.67107229950.50058419470.48963:69750.481747805611.21462430

TRAIN 7 0.20506328880 12 0.37869999600 12

PHY OM 2.0-56962025' 0.-4259611780

NEG WT 9 0.0 0.0

GED 10 2.581645570 0.1599516677

SVP 11 2.956329114 0.1386916749

6MOCC 12 0.57063291140-01 0.33045940530-01

(XWFOCC .3 0.4154746835 0.2300809522

2:8H1cc 0.10556962030-01 0.102013959270-01

DA /Pw 15 10.09615030 15.44776192

NYL/PW 16 1.925543429 2.300089160

AVEMPF 17 0.3171822765 0.2863453859

FOSL 18 0.31807681290-01 0891578830-01FOSLEX 19 0.67439290550--01- 0.67260813240-01

ATPFAS 20 0.41330379750-01 0.15993094460-01

MINING 21 0.4802458420 0.2010955900

EDS16 22 6.291139241 5.171336726

UNXMPF 23 .0.1942523207 0.2957062699

tFMOCC 24 0.4460316456 0.2366932389

MINWC 25 0.5030949367 0.2136431534

J,

358

Occupation Stratum 5 Black Females

rj . ea .4 -4 N -I ++ .4 .0000 C C 00 00 CIiii I

cloc)c) c) ,I -. oO LI ,S

'",.. ;?, 8 ;4 :1; :4' 4 gc, .3° -. ... 0 .0 TA .0 el eV Ur 0 CP .-4 C)J _ - , . _ -

- . N0 h. 4 -. ' 0 1 0 - 1 c0 C) 4. r4 43 C).0

...r. t.24 .17 1 .4'3 r...-5 . C

43 41 .4 41 MOW 04 0O .4 ra .3. 0104 N+ ... .4 .4 .4 1... 0 4. .4 v. PM el 1.

0 .4 v.

0 0 C./ C) 0 C) 0 0 0 C) C) 0 (3 0 0 C) C) C) C) C) C) 01111 1 1 I 1 I i

..4 .4 N.. .. .4 .0 ,JO op 0 0 0 C) C) C)Mt 4 I I

q.10 t'. -.:340.4.4440 TIN P. ,4 " C) 0 C) 0 (IN 1 w. el 4 r. C) 4 re) CP n. C)NN c0 P.- CP 0 0

P. ,a ,.. ea .4 .4 el 4) a.N

ra 40 44 4. UN 0 4' CP 0 r4 ..1. r4 47 0.4 -1 0 ..... r4 .4 4) 0 uN a) 4N ru N0

A i.7) r3.40 5 W.- 8 "'u1r... 40 c0 -4 .)4 0 it." .41 I.- 0ra 0 r. .2 0 4T a3 0 4. 4N .4 CO C7.4 .4 .4 .4 r4 .-T P.T. el C) Jr.. r4 rn r4

1! -: rl rl 1 -: .: . . . . -c, c, c, C) C) CP1111 C) 0 C) C) C) 0 C) C3 C) C) 0 C) C) C2

.4 .1 .4 -+

1 i I i i 1 1 1

.4 .-4 0.0 0 0 (7 0 0 c)t t 1 1 I I I

j!:

0 0 00 ..1 4 ,o .7 ri n4 0 CP C) n4 C) 0W 1 . 4 . .4 40 C) 0 C ) . 4 " .0P 43 0 0 UN P. 4.. eV eV C)

a) 4. 43 0 N ... C3 1/ .3 .0 4P N CP .1 0rd% ..4 CO U, ,1 F4 44 ra 0 C) al UN 0.4 ea P. /3 4. C) .40 .4. al ra en 0 4) N uN 4. ... MO," ro, NI eN .-. .. 4., fq 0 .0 ,n r4 .0.4 - 0 0 14 .40 0 0 0 0 . loot ... e .4C) 4) 0 0 0 C) C) C) C) C) C) 0 0

1 1 I IIII I I I

..I .4 4 . NI0 0 0 0I I I 1

C)4 ..t r4 rn 0

0 0 . C)4 ,.. 10 1% ... 0, a) CO 44 0 ul C)

PIN 03 P- C)NJ 0. 0 c ,

. 3 ° .1 7 . rN C) . 3 0 4 4 rn 4 .0 0 C)41

%Ft CO O r-ni

0O -.., 2. .4, .4' .74 ll, c..;

OPg

. - 0 40.4

D C) C 0 ., J'0 4N CP 4. ...r c,... 4 I rl .. . .4

e4 e0 . 0 43 r4 r.

C) C) C) 0 0 0. 1....1 ;,..

0 C) 0 0 0 0 43 C) C) 0 C) C)11( 1 111 111 1 1

v.

I ?0 0 0I I

0 0 ,.... 0 CIuN .0 r4 0 .0 0. ,N1 IN (NJ e- 03 0 04. el JO ..'1 44,7 a) 4J 0 .... 0. 0.4. e- c3 0 eV , .7 3'r .+I rn a3 P.- t- 00 0 .0 0 4 * Z. " ti P. C 0 N CO .0 .0 00 .... Ws 0 ,.. .4' ... '... 0 0. 0 CO MM 0v. at . . .+0 rg .4 .4 .- ra r4 0 .4. e,,, .

. .4 . .0. ....off .4C) 0 0 0 0 0 0 C) 0 0 C) C) C) C)

1 1 1 1 I 1

.... - ... - ._. .. 0.,

o .0 0 C) 0 C) C)I III fi

0 0 030 00. 0 0 . aD .. .r ,r, .1 .1 - NO4 MO NJ 0 r- -4 0 41 4.4 OuN rn 0 ., .4 Cr C) .4 C) Cr ).- V, uN 00.00 .4 4 -. 0 4 C) 4. ,1 0 r C)

e. e) e... 0 ... as 0 ,..1 fsj .4. NOP. r1 03 -4 4 .. U% ... NO P. 0 .4

.. 1..40 0 0 0 0 0 0 C) 0 0 0 0

1 i 1 1 1

0 0 C 0.. g1 A,0 00... 0 J,.... .0 P.- r1 .3° CO .- 0

MU .. 7 4-. 4 4 0,4 e4 0N 00 C , c , 4, 4. r. ...r 0 1.0 (0 40

0 0 1'1 0 -+ r. n) 4 0 '0M0 0 4 0 r4 ,r e, D co 0rN *3 .... .... 0 Nn ,r0

a. e ..C) C) C) 0 0 0 0 0 el C7111)111

)cO

QC 004 0

. 0 0 0 <3 0 C) C3 C3 C)z .4 . . . . ,

1.1 (7 0 C) 0 0 0 0 0 43

P. (4.. 04.1 .1 CC - `C... 0 .0 1...../UJ at 4 .4 "4 0% ...t. 4 4) P. 0 41 . .. ea r1 .1. us .0 .... c01 7 47 .4 eV rN 4. r. 41m ..... z - x -. -..4.4.44...+X X 7 X0 0C z C .0...) > >

ys4f4,11EhuKatA

17 id 19 20 21 22 23 24

I 4..1 T',>,,,,0 0.',h)1110-01 3.145601 0.111313 -0.175940 0.181068 0.154715 0.1517752 0.4,i. '3 0.1S3418 0.216115 0.1'6128 -0.1,5316 0.1508510-02 0.470E141 , -0.1197733 0.11:1,o) 0.45777 , 01 0.110313 0.176717 -0.102438 -0.1610110-01 0.702115 -0.1732744 -C.4c:,%60-0. -0.4411010-01 -0."..41400-01 0.6102440-01 -0.2637290-01 -0.154666 -0.408371,1 01 -0.9091660-015 -0.14,1 )1 3.19',,JZ4J-01 0.p51o340 -02 -0.1234133 0.918i040 -03 0.916816 -0.0201610 Oi. 0.1666960 -016 -0.013K120-01 3.7141b0001 -6.1281270-01 0.'668173-12 -'0.1161570-01 -0.161185 0.1165600.01 -0.2464457 -0.113464 -0.191)289-01 -0.153/114 -01 -0.1119100-01 0,116982 -0.4138810-01 -0.4814420-01 -0.3531030-01& C.51,4810-01 -0.1279280-01 0.148154 0.104866 -0.299618 0.174581 -0.4079810-01 -0.15112074 0.' 0.0 0.0 0.0 0.0 0.0 0.0 0.010 C.8'16090-01 -0.1671560-01 -0.196121 -0.316911 0.170767 -0.6819170-01 0.331,5760 -01 0.09832111 -4.1721164,-01 -) 1431890-01 -0.1771700-01 -0.152060 0.160559 -0.8288360-01 0.2161480-01 0.1212570-0112 -C.704/440-01 1.150;320-:.! 0.4215810-01 0.128112 -0.5750409-01 -0.130783 -0.129114 -0.73148011 0.3998600 -01 -0.718160)-0) -0.182192 -0.141419 0.151272 0.3361920-01 0.1439910-01 0.9,044514 - 0.369259 -0.953/310-02 -3.140447 -0.53411 0.308393 0.111115 -0.180503 , 0.66066915 C.671823 -0.142104 -0.108416 0.100806 -0.23256) -0.126938 0.496814

s0.301748

16 C.694149 -0.6217600-01 0.108192 0.713234 -0.411100 -0.194919 0.437177 0.1645870-0117 1.30000 0.6224970-01 0.130663 0.592817 -0.558036 -0.7800480-01 0.749664 0.2194170-0118 1.00000 0.623152 -0.1609580-01 -0.141136 0.8738230-01 0.1623800-01 -0.701767.3 11Iv 1.1 .,100 0.300802 -- 3416.5 0.8463820 -01 0.14a88. -0.18115620 1.00000 -0.657822 -0.7114460-01 0.391103 -0.15505321 1.00000 -0.3933130-01 -0,347218 ' 0.16617622 1.00000 -0.3897450-01 0.3790570-0123 1.00000 0.2133570 -03.24

f 1.00000

121.1 14131f 25humt10.

C.14 ":e."2 -0.1427513 -0.16014;4 -C.821c0-015 C.348514i>-02

7

8 -0.152118

10 C.,:771211 0,53241,0-0112 -:.457.'111ik 0.64t,Is C.2./71,6lb 47.4745,.C3 -03

17 C13

19 -Z..1'4616120 -0.114418

0.1,75114

22 C-21766%,-0123 4.3474100-0224 C.994456

25 1.00000

C)

0C't

360

Occupation Stratum 5 Cross Race-Sex

VARIABLE 11EAll STANDARD DEVIATION

SCHOOL 1 9.687727825 2.627930192WAGE 2 2.480206561 0.9147806731UNION 3 0.5212636695' 0.4998514205MIG<50 4 0.5030376671 0.5002948107S 16 5 0.3037667072 0.4E01627809EXP 6 29.69380316 11.61176717RACE 7 0.1117861482 0.3152947765SEX 8 0,3888213852 0.4877790632TRAIN" 9 0.1081409478 0.3107471877PHY DM 10 2.077764277 0.4229186710NEG WT 11 0.1701093560D-01 0.1293905182SVP 12 2.927339004 0.1714929166MIMOCC 13 0.6446901580D-01 0.4253584376D-0114FOCC 14 0.3444240583 0.1902614573,DFOCC 15 0.2696233293D-01 0.8365-155980-02DA/PW 16 24.00595430 55.95157687AV'EMPF 17 0.4221124342 0.2803493161FDSL 18 0.3459802204D-01 0.4481312654D-01ATPFAS 19 0.4773171324D-01 0.1667134298D-014flININ 20 0.3475637908 0.1746140554EDS16 21 2.772782533 4.497702891(INXMPF 22 0.2c-'0893884 0.31.81097695XFMOCC 23 0.3, 3803)13 0.1942210881iMINOC 24 0.4358554070 0.1707313442RAXSEX 25 0.3402187120D-01 0.1813956022

0'.1..4C..)0 fr:3

C08REL1TI08 nA78IX

TAVIAdLEitOWIER

1

2

3

4

5

67

8

1

1.00000

2

0.2435561.00000

3

-0.259644D-010.2767801.00000

-0.09110.'3-C2-0.8491710-010.5822340-01.00000

5

-0.140773-0.181035-0.165631-0.8856240-01

1.00000

6

-0.469836-0.1028300.4576770-010.6026190-010.1105350-011.00300

7

-0.855559D-0t-0.130286-0.1510070-0-0.715663D-00.235227

-0.4347300 -C1.000G0

0.7490010-01-0.4215710.1786490.2008000-01

,-0.2279270 -010.1353520-010.614748D-01

1.00000

1.181.18L2 94;mnd

11 12 13 14 15 16

0.211232 -0.1,,S451 0.6573000-01 -0.148,,,67 -0.293755 0.235451 -0.2648160,01 0.3112740-01

0.145712 -.7.4994)10-01 0.4447400-01 -0.4641450 -01 -0.5205443 -01 -0.6307700-01 -0.167245 : 0.230333

3 -0.1090b00-01 0.95/5480-01 -0.4321920-01 0.3933390-01 0.4416040-01 -0.114300 -0.9014110701 0.866676D-01

4 -0.3732890-01 -0.4711540-01 -0.383855D-01 0.1391920-01 0.9422730-02 -0.8384260-04 0.1121810401 -0.277507D-02

5 0.1671550-01 -0.402641D-01 -0.4602820-01 0.3029620-01 0.6717150-01 - 0.4489183-01 0.2067300-01 0.5777530 -01

6 -0.275306 0.7719080-01 0.7472510-01 0.4312443-01 0.173597 -0.121620 -0.4224800-01 -0.2329760-01

7 -0 1178303-01 0.4420493 -01 -0.4666860-01 3,3565680-01 0.136596 -0.8604690-01 0.3941840-01 -0.3647760-01

8 -0.7709033-01 -0.105469 -0.104925 -0.3124570-01 -0.301677 0.435640 0.187836 1/' -0.123942

9 1.00000 -0.5461010-01 -0.1555100-01 -0.9888590-02 -0.114195 0.123937 0.1100840-02 0.107470D-01

10 1.00000 0.2025963 -01 -0.2767130-01 0.446536 -0.395830 -0.232645 -0.9824770-01

11 1.00000 -0.262215 -0.2753410m01 -0:218518 -0.363547 -0.1773050-01

12 1.00000 0.204105 -0.4573500-01 0.448276 0.3730110-01

13 1.00000 -0.641743 0.874430901,02 -0.1145683 -02

141.00300 0.456257 0.1790020-01

15,1.00000 -0.427412D-01

161.00000

a0U

00000'1.

SZ

t0-130POP9f'0

Az

tt-c^t,Cza"0

IT

1C-C199StS*0-

ZZ

LC-COCOPDR.0

IT

tC-C9 0LCcL*0

CZ

1.0-CV'SZtt.'0-

6t

29-c:c.atP*0

St

10-Ct*F6CP7'0-

LI

tc-c(rIt511-0-

91.

IC-CL7t-ft.-0

St

to-co,FrZ9Z*0

hl

1u-c94CW-0

Ct.

70-cttO9tS.0

TI

tf-(707.0..

Lttc-c,,77cf.otw:.

Ot

CO-Cl^ttCQ*C-

6

tf7.Sti'0

O

400(TC0

L

to-ciACt,:*0-

9

t0-Ct9SE9t-0

S

t0-0PtCt,11*0-

t0-n96iZ;f7'0-

1LC.Pk*t*C.,-

1r-cot-s9m.0

C2CUCX

SZ

VISYISTit

006.'0'1

ist

0E0186'0

00000.1

Et

10-0081PW0- 10-0ALS9LE-0-

0000J'1

tz

Z0-Gt6t6CL°0

Zu-Oti.tteL°0

10-0ttZtltS'O-

C0000"1.

tt

t0610F°0

'

PbtZ62°0

616SEE°0- 10-0099SWO

00000.:

OZ

tellS9Cu-

tE9SW0-

969I000

1.0-008EEt8'0-

EtAten*0-

00000't

Ft

SOLOt°0- 10-ObS6L60'0-

99srozo0

t0-0C69tht0-

09Z6SZ*0-

S''S[Pt.0

0000Vt

Pt

EL9901.*0

EPULt°0

hOSCS9°C

t0-Ot?ElSZ7"0

ZOott1-0-

tS96W0

6tILSSI0

00000'1

LI

L0-OPLQSMO

t9-04,169St°0

7s6t[Z°0

10-C2'0901C0

6LE,4Pt*0- 10-0ESZZ0- 10-0tSittS°0

1Zt0Rt"0

Qt

LZ96SS*0

PZ006f) '0

F7tFtt°0- ZO-COZSIS1'0

LOZS,WO

ei1tSP1.0- t0-0007tFL°0-

SSL1191'0-

St

Lt,91.c'0

597666.0

10-0a0tFt°0- To-ocs/66co

0C7t6°C

Lo*Ot°0- t0-0S1Z,PP°0-

t9CC:71'0.

*1

SLSSiabe-

tozervo- To-oarPoo

ZO-CitP66S*0-

Pt5E71°0- CO-OfmStg°0- 10-CL40t:t"0- 10-0,7tS10,00-

Et

10-CLP*,Stt'b

I-001,61)SZ°0- to-chr(rito- t0-019069[°0- LO-OSPOt/t9"0

1.O-ctotc.60- 10-0i4'1ZtS*C- to-cr1 at-90- Zt

C91pc4-0-

7ZL6ZZ°0- 10-0S9St1S'0- t0-041(164-0- 10-491(opnt'0-

tO-Ottr(07"0- 10-Citcp,,,rt.0- to_ccl.rcLie_

It

^Si:int-0-

10/L6E-0- to-OrICOR-S°0 .10-0761t0-

*Thortt C-

7,017W0

SttOtt'D

10-0ScI1WO-

Ot

9t/6010

Ast.'ziol

To-crsziro

to-crnhcco-qo

1.o-c59Pktz.c- 10-naffttl."0- t0-0t,tZLF'0

10-CLTL,,71*C

6

Ot SFI tO

6hAtqh*C1

SeZ091'6- 10-CtPtiPt°0

St^0Stt°0

6r6itt°0- t0-CePtCL.0- 1.0-19Sc4W0-

9

t0-06S766S°0- 10-0016SZ14-0- 10-089.779(0-

ZOZtql*0

t0-CiROI,Pf-O- to-aS459sz-0- t0-eel/v.1C0

tt-c,ctr«c-0-

L

10-00tcEp6°0-

1960710- 10-C6ttiOi"0

t0-01.6bLF6*0- 10-0,16tf,..h*0

tfLttt'0- to-CAS6ttt,'D- 10-tt[t1ft?-0-

9

10-Cf.L9tiC0- 10-0t9L0E9°0- t0-0,1ZSEICO-

16pfr60

t0-05579990

t0-0,4771.6'c- t!)-(1LCZC(t*7,- iC-CtrPt.4tC-

S

Z3-0La03Z*0

rp-ooclohou

10-00PO161°0- 10-07/6tSCO- 10-C9c7nt,C0 10-66"0

10-0719609*9- t0-C,(qi,4*-

t

n6LOZt*0-

C4S,,tt'0-

POIF6SL°0

tfOII°0-

Ctit0./7c-

t6ttici'0

(4CIZI'C

trOf77C

t

10-CSI,S6t6'0- 10-0$,I)6Fp0.0_

OSC9W0

9LttZt*O-

9trtt,1*0-

Sttitt-0

ciGtwt*0

Cf*.t/f*0

t

F6LLPt°0

tItt,6ZZ*0

tO-OSS5fSti0

10-021t4LO°0

10-0t,CSI,t9°0-

toCPW0 tO-C,SCZS.0

tCZsPt°0

'

t

qwnot

Et

ZZ

IZ

OZ

61

at

It

2111118ti

VAR(A8LE

SCHOOLWAGEUNIONMIG<50S. 16hXPTRAINPHV DMNEG WTSVd%BMOCY,WFOCC%eFoccDA/PWAVEmPFFOSLpiPI:ASZPININ[psi('1VvXPFFVOCC

%MPIOC

A

363

Occupation Stratum 6-9 White Males

- __MEAN STANDARD OLVIATION

1 9.590645161 2.6484252712 2.957385305 0.98016378813 0.5591397849 0.49/38233324 0.4597813620 0.49919354275 0.3046594987 0.46109013256 29.75985663 11.936117497 0.1541218638 0.36171434268 3.043010753 0.7621530251

1.161250323 1.20212'0910 4./87313620 1.15031832711 0..3205/347670-01 0.46425196370-0112 0.57648745521) -101 0.124091048913 0.49569892470-02 0.78935',97350-0214 32.3337572 77.18030955

15 0.3/32497013 0.2682779131

16 0.29120546440-01 0.35846870)90-D117 0.44227518570-01 0.16032933450-0118 0.2515325161 0.14083?967719 2.63082430 4.241358082

20 0.2.4547264C4 0.30012138)21 0.62605734710 -01 0.130229358122 0.1446630824 0.1224613621

C1,1AtL1TICN 1ATP1X

vA21A0LEMd,88,1

1 2 3 4 5 6 7 8

1.00000 J.25542.7 0.9945110-0 0.3440240-01 -0.216698 -0.450620 0.214152 -0.7070800-02

2 1.0.Mbo 0.423892 0.363224J-01 -0.1'5)97 "-0.118761 0.100352 0.4519250-01

1 1.00000 0.160062 -0. 06482 -0.1466740-01 0.391280)-01 0.9764540-014 70000 -0.109339 -0.2913570-01 0.2534810-01 0.4249560-01

S 1.00000 0.2837380-01 -0.131571 0.1570616 1.00000 - 0.233012 -0.5105490-017 1.00000 0.1501230 -01

8 1.00000

10.4101A81E IC 11 1. 13 14 15 LaNuw,Eu

1 -t..150i213-C,1 0.616100 -01 0.1491260-01 -0.91457v1 01 -0.1125100-02 0.125968 0.7315370-01 0.9328180-01C.100," 4 3.1,11,,t, -0.1,.549.. -0.%481174-01 -0.111.041 0.107939 0.151174 0.1846850-01

3 C.231,1"2 3.12952) -0.119475 0.1fe7610-01 -0.5707100-01 0.3/10420 -01 0.255555 0.4861800-014 -0.3343e50-01 -0.4154540-01 -0.2123323 -01 0.114684 0.2602210 -01 -0.7025160-01 0.1797820 -01 -0.1274690-01

0.177103 -0.2950730-01 -0.110214 -0.104244 -0.127834 0.5823140-02 -0.187443G-01 -0.1006396 C.8493820-01 0.4424470-01 -0.1171640-01 0.7650240 -01 0.7418590-01 -0.103540 -0.3322360-01 -0.8296800-01

-0.2428390-01 0.4516230-01 -0.4415140-01 -0.2403370-01 0.4649750-02 0.6345200-01 -0.3996491-01 -0.3217020-010.604873 0.113431 -0.0044760-01 -0.364054 -0.542603 0.6301910-01 0.9188930-01 -0.1774880-011.00000 0.449626 - 0.318569 -0.170175 -0.339503 0.9267240-01 0.313317 . 0.7481100-01

1.00000 -0.496038 -0.3136110-01 -0.9794830-01 -0.5289620-01 0.1266411' -01 -0.104050II 1.00000 -0.347068 -0.162800 0.133902 0.8465000 -01 -0.1060870-0112 1.00000 0.765068 -0.8816070-01 -0.125579 0.3073900-0113 1.00000 -0.131278 -0.125043 0.9864220-0114 1.00000 0.455112 0.232623IS 1.00000 0.29Y60816 1.00000

VA4LAULE 17 18 19 20 2161,018E R

c .t '4 %I2 -3.:213411) -C2 -r.Ael14c00-01 0.119554 -0.9341730-01 -0.9368980-012 C.,,r',:r..,-,:i -3.1-)12. -o.,,e141e1) -(11 0.4.-05109 -0.6222030-01 -0.1289663 C.111.-, -1.(5...17 -0.143410 0.13s 16 0.9191110 -02 -0.355131-60=014 C. ,4.)40-)7 -1.4450120-01 -C.9351,;.'10-01 0.118812 0.110856 0.1064065 -0.t204c1 -0.114149 0.014767 -0.107073 -0.107079 -0.1556536 -0.4113"5/-02 9.2422,'70-01 -C.415',r10-01 0.2521660-03 0.00:9880-01 0.5711840-017 /t.')2 3.015x5 -01 -1..511550-01 -0.124562 -0.9711000-02 -0.2260690-01 -0.4261500-018 -C.95/4740-01 -1.14:,c7 ...Loin,: C.763a100-01 -0.3197S3 -0.4381629 C.9 "51 5 70-01 -5.155954 0.110,103 0.359667 - 0.132120 -0.315060

1 0 0.,2471 1.3451/6:)-01 -0.7042470 03 0.6614380-01 -0.3583690-01 -0.226161II -0.2Sr5C4J-01 -0.495)15-51 -0.14250 0.5139010-01 -0.340577 0.16920700112 -0.162:773-01 0.458/79 -0.9449570-01 -0.117316 0.909238 0.931048I3 C.110122 0.453632 -0.114581 -0.124959 0.759419 0.77798914 -C.171129 -0.1911E4 0.1701550-01 0.267411 -0.9196240-01 -0.4703350-0115 0.348564 -0.2421/5 -0.1173160-U1 13.720508 -0.127239 -b.10322016 C.172754 -0.153579 -0.9342200-01 0.213956 0.3526910-01 0.3346450-0117 1.00000 0.6649150-11 -C.7942370-01 0.244231 -0.3109240-02 -0.1462600 -0113 1.00000 -0.8234050-01 -0.288295 0.464650 0.47532419 1.00000 -0.7742470-01 -0:9698670 -01 -0.14341820 1.00000 - 0.119360 -0.10744921 1.00000 0.9343192? 1.00000

I

365

Occupation Stratum 6-9 Black Males

yARiAOLE

SCHOOL 1

WAGE 2

UNION 3

M IG <50 4

S. 16 5

6---EXP- T

MEAN

8.8494623662..359516129

0:43548387100.47311827960.698024731228.263440E6--

STA41DAkD,DEVWION

2.9406:)08430.94179029180.49715844690.50062442850.459963482411.86092640

TRAIN 7 0.1075268817 0..3106180619PHY ON 8 2.903225806 0.6666119574NEG hT 9 0.7096/74194 1.153694886SVP 10 4.330645161 0.92723163261:8MCCC 11 0.1076666667 0.3744345942D-01

--%WECC0 12 0.38602150540-01 0.5343015441D-01-.%FFOCC 13 0-51720430110-02 0.662856GU63D-02DA/PW 14 21.80380573 143.42801756AVEMPF 15 0.3162598566 0.2689391596FCSL 16 0.472460(: 94F0 -01 0.1094640047ATPEAS 17 0.39948924730-01 0.16611o8886D-01gv1141N 18 0.2E20625306 0.14;3201:',648

ECS16 19 5.7419354E4 4.470693098UNXMPE 20 0.1923949821 0.289167-'161

21 0.41774193550-01 0.5933',1'6930-01tVINO0 2 0.15144C8602 0.55205351D-01

:1,83

VAKIA6LE

ALItitA

MATRIX

23

45

67

-

I.ccac.3

0.19

4143

0.1.30544

0.3028190-01

-0.329417

'- 0.4596..L

0.207185

0.4214990-01

1.o0oc0

0.4/7591

-0.25''19

-J.274858

-0.2340930-01

0.3639519-01

0.0000450-01

31.00000

-0.22

7-0.274505

0.124357

0.1016221-01

-0.51:4590-01

41.1

JO

-0.8228030-01

-0.9119910-01

-0.508330)-01

0.8931970-01

51.00000

0.100817

-0.112087

0.1023323-01

61.00000

-0.216070

-0.130032

71.00000

0.180998

a1.00000

v6k1

.1.8

1(61

.' "E

A10

1112

13

14

15

16

1-0.4743'01-v1

-^?.2(.,G,.154.6-01

-C.1),a8e1

6.81v:16J-01

0.6501180-01

0.8606180-01

0.1443120-01

-0.3259110-01

2.

C.112:,..1

3..11412

-6.159510

6.874404J-01

0.4109k.50-01

0.5102720-C1

0.346695

0.112910

3C.1111S9

0.2757e6

-C./21436

0.4105280-01

0.5195410-01

-0.3734770-02

0.411086.

0.9221843-01

4-C.1011S6

-0.193264

0.111464

-0.53454,0 -01

-0.11[022

-0.1849120-01

-0.8012690 -01

-0.189384

S-C.5355060-01

3.5274C67 -02

0.5317450-01

-0.5148713-01

- 0.5915410-01

0.2914030-01

-0.125154

0.2159280-01

6C.3721460-01

-0.1428650-01

0.9441140-01

-0.2210420-01

-0.1130500-01

- 0.7816880 -02

0.030004)-01

0.47/9000-01

7C.102649

0.104858

-0.120591

-0.7362183-01

-0.119247

-0.4276000-01

0.6133331-01

0.8070370 -01

8C.412900

0.205000

-0.4115610-02

-0.303238

-0.594231

0.326716U-01

0.8408910-02

0.6297580-01

91.00000

0.634824

-0.404.942

-0.9876440-01

- 0.250618

-0.2942640-01

0.339992

0.254176

10

1.00000

-0.503993

0.3542360-01

-0.7245130 -01

-0.3011910-01

0.239844

0.349893

11

1.00000

-0.426921

-0.334059

0.139733

-0.209031

-0.225793

12

1.00000

0.818713

-0.162028

- 0.217329) -C1

-0.8707820-01

13

1.001100

-0.166115'

-0.425247)-01

-0.151410

14

1.00000

0.333302

-0.3723550-01

15

1.00000

0.138210

16

1.00000

N/A.1Antt

C19

20

21

'22

M..$06ER

1c.se45490 -01

6r.J-01

0.1,5738

0.5710400-01

0.0408160-01

-0.1906690-01

2c .2

7(44

3-).131105

-0.15..509

0.456465

0.8333030-31

-0.1856530-01

38.5,)81.4

8.5,)81.4

-3.1234E3

0.159569

0.4405891)-01

-0.9940170-01

4-C.2346660-01

0.60ril2p-q1:=1

-0.107488

-0.8118110-01

0.9819400-02

5-C.737)66

-1.9 34710-01

0.845231

-0.176426

-0.5035290-01

-0.2667570-01

6C.47c31.6,)-31

-0.61.321.3-01

-C.94w227o-1

0.121921

-0.2116590 -01

0.4430080-01

70.1919/1)-01

0.1.7451

-0.4218950-71

0.2134100-01

-0.7961670-01

-0.167327

>-0.1946700-01

-3.2c7111

0.4)8),,,0 -01

0.3282950-02

-0.393440

-0.425093

9C.110611

-J.16a139

-0.1451450-31

0.253422

-0.110932

-0.400196

10

C.1154.15

-3.110455

0.157'.01D-J1

0.230517

0.23a020.0 -61

-0.316069

11

-C.17e213

-1.6145C50-1

0.1311733 -01 -0.137822

-0.421948

0.224351

12

C.2C641.

9.52tAts

-0.1701170-01 - 0.1129593 -01

0.9,4577

0.783943

13

0.195624

0.422345

- C.'75(.040 -01 -0.4230870-01

0.932917

0.742115

14

-C.73(f1130-01

-311169759

0.82060-01

Q.189379

-0.1646/3

-0.8227900-01

15

C.52t126

- C.2430'6

-0.8864140-01

0.76/123

-0.2431900-01

-0.168304

to

-0.246925

-3.216967

-0.8600133-02

.0.096,040-01

-0.9532040-01

-0.255522

17

1.00000

-0.3989690-01

-0.145557

0.506886

0.237758

0.102505

18

1.00000

-0.7546433 -01'-0.259967

0.338791

0.322788

19

1.00000

-0.116704

-0.3917910-01

-0.3322090-01

23

1.03000

-0.2390140-01

-0.153015

21

1.00000

0.768844

22

1.00000

a rY O

367

Occupation Stratum 6-9 White Females

vA/TAltF

crqint7

E AN

n.16949n566.1.°3f,'037740.4'330672642

STA(,!0A20 DEVIATION

7.66757547B0.4R072541771.49717413.33

"fle90 (1.'49°679745 0.c079174931

C. t4 r. '1.749?n3n189 0.4122913551ryr). .31.90041396 10.74941753TDAlm 7 0.7win1rFA790-91 0.16/4,217370

WiN0 0" 7.1017,1595 0.A111901756'on wr 0.'(.41g00414 1.777156886SVP In "3.19(10413/6 0.7814915335conAncr,

m!W="Cr.

11

1

n.11r:6(-n1774D-11n. A$111(1/0157

1.7910161777D-010.317316x414

°"Icr1CC 13 0.3780672642D-01 1.19753503470-01nApw 14 9.61)276917 16.9391:i156

0.1617443196 0.1907(130512

1, 14 r).?,132qconsin-01 0.1070 -160047D-01

ATPFAs n.,:i44r8A7070-01 0.1?7)9-190130-91'%iTmIN1 0.61173320?7 0.1809724475

cncl4 In 9.41,'264151 4.41114202Sowc4or /n 0.5(177'10037D-01 1.12'7317796m,rk4ncr 0.77612.11321 0.1925912933Tmtmnr 27 0.7461471609 0.12/8800414

(191-t11109 !Tot(

vall3niePi.grO

1

1

S7

1

1.1W,10

7

-3.1461421.0npoo

1

-9.1483040.1169271.00900

4

1.1)1143-3.200216-0.2997401.00000

5

0.146181-0.119773- 0.3664450.2235610-011.00000

6

-0.3502100.4679540-010.299936

-0.1067340-01-0.244412

1.00000

7

0.1057160.117945- 0.1494310.5046390-0t0.349250-01

-0.9853040-011.00000

8

0.2298130-010.6192593-020.6858230-02

-0.111399-0.9414340-010.8475760-01

-0.116711

Vic!IVALF4,1",r2

A

5

7a9

111!

121114IS

".117,7 10_11

-11

-''.1 )97,°7,_11.0.c179410-11n.o117^94-11

0.7441541.9'1119

In

1.161701-1.1641010-1?-1.1:5442-1.4174771-011.°14a100-01'.4Q1,..673-011.1474191.4716561.9560,4wyleno

11

0.,714551.4774700-01

-9.211,260.4481420-01-0.4641950-01-0.741491n.1192730.3118660.1891780."91576I. 00900

I?

-0.?77174-6.41n837)-01

=1481-010.6510620-010.706962

-0.784550-0.445292-0.640928-3.577357-0.8913651.00000

13

-0.1'1619-0.7166150-010.171127-0.4642950-010.5485850-010.151553-0.148197-0.345162-0.629266-0.388618-C.6700280.8051371.00000

14

00:11=-0.144582-0.7225150-010.141742-0.2242550.103910

-0.1260000.6932340-010.1688690.5801890-01

-0.9492740-01-0.244428

1.00000

15

0.2102430.150111

-0.2485100.3179720-010.1655210-01

-U.1631840.2378190.3519580.6355460.5217200.587862

- 0.748560-0.6504990.4969001.00000

1.0000016

-0.1911920.1129780.152245

-0.114568-0.1256270.101601

-0.2040710.2042970.171829-0.8121693-01-0.2779340.2168960.3128910 -01-0.6392960 -01-0.267609

/ADItAIFNormc*

17 18 20 21 22

1n.,C1.^7 -1.1491,4 9.7671.41 -0.6111570-01 -0.223997 -0.216615

9.19&7n19-11 -1.Inn'014 -0-1(4797 0.110159 -0.4440240-01 -0.349:020-01

1 -'."1151 1.271"1 -0.111SiQ 0.4608)8 0.230136 0.233483

e. n.114,11 , -1.651A/41-11 0.715944n-91 -n.182168 -0.4581960-01 -0.4545430-01

S -n.,90,419-01 -1.119°4" 0.967116 0.6107361-01 0.6845250-01

_90 .-4,,4 1.117091 -0.116105 3.189973 0.236057 0.197461

7 0194975 -9.1c,160% n.6411165,7-01 -').7920763-01 -0.740115 -0.272149

4 0. -./. /-,/ 2 9 -1.27c/21 -0.9117940-n1 0.360256 -0.446530 -0.451740

n 0.'.44447 -1.470744 ..0.414,.1,0-,1 0.507513 -0.646153 -0.661478

11 6.1418,8 -1.471,80 0.1 n}17 0.316471 -0.575185 -0.539777

11 9c...9949 -1.371050 -n.1347390-11 0.201350 -0.848388 -0.869636,

1) _n.474116 9.4970/.9 0.4273850-01 -/.101171 0.1)9)12 0.998435

11 -n.c184,17 1.5517c,3 0.cS51543-01 - 0.341184 0.823445 0. 28079

14 r.110941 -1.410471 0.21149n 7.1537081-01 -0.111816 -0.106667n.4,415q -1.597P54 n.4117811-01 0.1634750 -0.751711 -0.756927

t. -n0990114 1.12134? -0.117069 0.2091873-01 0.201164 0.201107

1.-11119 -1.643499 -0.1865270-01 1.111858 -0.434170 -0.642637

18 1.00o00 -0.206564 -0.274215 0.514303 0.521274

1.01000 -1.174954 0.4341570-06 0.4553290-011.30000 -0.306464 -0.312347

71 1.00000 0.9992261.00000

1.60000. _

VARIABLE

369

Occupation Stratum 6-9 Black Females

MEAN STANDARD DEVIATION

SCHOOLWAGEUNIONMIG<50

1

234

10.116279071.721860465

0.39534683720.4418604651

2.3725544030.47812259740.49471179120.5024855166

16 5 0.6279069767 0.4890834876.S.EXP 6 28.37209302 12.09687731TRAIN 7 0.2325581395 0.4274625744PHY DM 8 2.162790698 0.6521113168NEG WT 9 0.2325581395 0.6109013365SVP 10 4.372093023 0.8247151258UMOCC 11 0.2511627907D-01 0.3197857118D-01'AWFOCC 12 0.6157441860 0.3513592039%8FOCC 13 0.4020930233D-01 0.3433487616D-01DA/PW 14 4.801214906 5.686813679AVE3PF 15 0.1333108527 0.1353513931FDSL 16. 0.2840904454D-01 0.3100309167D-01ATPFAS 17 0.3459767442D-01. 0.1141368202D-01UlININ 18 0.5865936475 0.2011410698EDS16 19 6.232558140 5.277252213UNXMPF 20 0.6325813953D-01 0.1258360268ItFOOCC 21 0.6559534884 0.3673662992%MINOC 22 0.6810697674 0.3403461400

S7

.'CORRELATION WHIZ

VARIABLE 1

sussza

1

2

345

78

VARIABLEsus888

1

23

456

89

10111213

1516

VARIABLEsumm

1

23

4

5

6

7

8910111213141516171s19202122

1.00000

9

2 3

0.2462181.00000

4 5

-0.4009690-01 -0.104038 -0.1054570.503139 -0.106571 -0.3635171.00300 -0.4900370-01 - 0.263175

1.00000 -0.2838871.00000

10 11 12 1)

-0.150519 0.112430 0.221056 -0.295434 -0.356303

0.335958 0.442916 0.202690 -3.341877 -0.477512

0.161228 0.384500 -0.2103510-01 -0.7145410-01 -0.167589

-0.187595 -0.6146283-01 -0.401858 0.351555 0.3067320-01

-0.101927 -C.250666 0.9417200-01 -0.4324150-01 0.6855170-01

0.6655813-01 -0.2399330-01 -0.6246330-01 0.7589340-01 0.278693

0.614939D-01 - 0.129736 0.344589 -3.281454 -0.199607

0.7394371.00000

0.424801n,5697361.00000

17 18

0.3336030.2806490.1388221.00000

-0.375415 -0.4834730-01-0.525943 -0.135186-0.373842 -0.20851.18

-0.866494 -0.3046251.00000 0.427965

1.00000

19 20

6

-0.576446-0.1565700.1331750.7560730-020.1487131.00000

14

0.2)12020.4631850.191797(1.1090663 -01

0.1045950-02-0.2092870.2188120.2366850.512)020.4645730.167727- 0.416108- 0.313316

1.00000

21 22

0.357159 -0.251765 0.245001 0.149282 - 0.313862 -0.320168

0.595577 -0.371767 -0.284964 0.471517 -0.371610 -0.382067

7 8

0.183992 -0.1048610.115494 0.257880

-0.107353 0.9096710-01-0.157250 -0.2247440.8210370 )1 -0.106167

-0.307214 0.1628450-011.00000 0.3178210 -01

1.00000

15 16

0.1556990.4131000.1613670.265220D-0

-0.171836-0.1115610.2629430.1507160.3015120.3544460.478609

-0.634201-0.3860780.5605701.00000

0.292465D-01-0.1237380.3486870.202748-0.291625-0.152559- 0.137345- 0.185211- 0.261505- 0.217022- 0.2809130.3234730.2639730-01- 0.141276- 0.138079D-01

1.00000

0.532314 -0.271119 -0.236694 0.624037 -0.8400330-01 -0.9264930-01

-0.130173 0.219195 -0.326996 -0.5806480-01 0.339852 0.329074

-0.204036 0.7426340-01 0.919912 -0.313065 -0.3495030 -C1 -0.2887673 -01

-0.194613 0.1163460 -01 -0.4d7565D-01 -0.6214546-01 0.9663380-01 0.100595

-0.5893550-01 -0.9183020-01 0.133774 -0.9891920-02 -0.287854 -0.273329

0.247,68 -0.321029 -0.170392 0.154607 -0.367601 - 0.365225

0.409845 -0.562535 -0.201809 0.340913 -0.515661 -0.521176

0.548194 -0.55:,578 -0.256141 0.537394 -0.377'046 -0.396755

0.349061 -0.2h5848 0.143038 0.233327 -0.057397 -0.831507

-0.544160 0.553492 -0.117422 -3.422437 0.996426 0.9)4118

-0.559519 0.6729490-01 -0.7530650-01 -0.260153 0.502779 0.513385

0.533105 -0.517700 0.3508590-01 0.316964 -0.427232 -0.445391

0.626880 -0.434050 -0.171909 3.663343 - 0.642838 -0.646403

0.14o039 0.136249 -0.259152 0.295505 0.312032' 0.310410

1.00000 -0.599430 -0.115060 0.719715 -0.572763 - 0.585437

1.00000 0.349300-01 -0.494802 0.540446 0.556494

1.00000 -0.358457 -0.119344 - 0.115379

1.00000 -0.428345 -0.4434281.00000 0.998830

1.00300

371

Occupation Stratum 6-9 Cross Race-Sex

v/01ARLF

Sr1-11-.11 1

w1 ,r3

m1Ge51 4.

S 16 6

6

MEAN

9.5130023642.6('07801420.52nr9456260.4860976359n.311146916229.96690307-

STANDAPD DEVIAT104.

2.6357362200.99920376190.50018763360.50042277620.466698837311.51194894--

p AGE 7 0.89P14515370-01 0.2862831.281

SFX 8 0.26713947q9 0.44298973160.1276595745 0.3341055222

PAY 04 11 7.787734043 0.8215853704MFG, vT 11 0.9010732861 1.191438200SVP 1? 4.A4i626478 1.071595136Y.rl'Ancr, 13 0.6733P96927D-01 0.51158150430-01TFlcC 14 0.2216312057 0.3426666837ev.111CC 16 0.1?510638300-01 0.11524296150-010A/Pw 15 21.1f-4)5290R 64.49889744Al/PmDF 17 0. 31812 ?2 ?22 0.2657181447FOSS 1q- 0,2043917050D-01-- 0.3955346660-01--ATPFAS 19 0.4162624113D-01 0.15522860280-01.wmiN1m 21 0.3568197721 0.224 827782Fns16 21 2.891257955 4.479232916liNvinp 2? 0.100(1367218 0.2715906152`1'F '1C(: 23 0.2351418440 0,359822084414r/V1r. 24 0.3024728132 0.3281326003RAXSEX 26 0.16548463360-01 0.1277230486

X89

C",

99cl

&T

IOn

mAT91x

VaqIARIF

I

por.qc*

7A

45

67

8

11.109)1

1.156100

-0.2311590-01

0.5447530-01 -0.110003

- 0.424602

-0.7032530-02 - 0.1820320 -0!

7-0.4900571-01 -0.157169

0.411313

-0.0722780-01 -0.154321

1.00n0n

..

11.00000 --

'

0.2708090-01 -0.255140

-''

0.5608440-01 -0.2918480-01

- 0.463264

-

1.00000

4-0

.481

4160

-01

-0.2969140-01

0.7433550-01

0.3113990-01

61.00000

-0.6330680-01

21)0:11626h

-0.1,804120-01

61.00000

120-01

0.6028630 -01

1.00000

-0.5888270-01

41.00000

16

vaotaAir

010

11

12

IA

1415

vp...,1-0

11.Ingli9

20.1,0'244

11.7716/90-01

40.741745n-01

S.41.4434510...11

4-0.716716

70.7401391-11

6-0

.134

699

9I . A no1 n

10

11

-......

1713

14ISI&

10117170-91

1.7619?7

0.1.70,01

-0.6744700-02

1.101153

-1.4953300-01

1.109)230 -0l

-3.007c73

0.5602110-01

1.00000

--.

""

-1.11458.1U-11

(rA,T41v

.'14 4

-a.3520560-01

0.116405

0.7381210 -31

-0.1024410-01

-0.330740

0.1941110-01

0.652506

1.00000 "---

00-,04790-01

0.5)11760-01

3.227729

0.104611

=0-01

-3.55,1789-01 -0.17,41110-01

-0.12601410-01 :0.1140150-01

0.2075333 -01 -0.110649

-3.6514550-01

0.163001

-0.171919

-0.460188

0.7966390-01

0.5561250-01

0.217413

0.265253

0.510369 "---=0.5623460-02

1.00000

-0.242591

1.00000

-0.104683

-0.1496.)1

-0.213959D-01

0.4594180-01

-0.6502310-01

0.118268

-0.5176720-01

0.806223

-0.161241

-0.575252

-0.410970 '-----0.462026-----

-0.238545

-0.664956

1.00000

-0.5706670-01

-0.1855S2

-U.5d19tIOD-01

0.1401010-01

-0.8064240-01-

0.108879

-0.8065500-01

0.776591

-0.135377

- 0.611055

-0.236240

-0.582745

0.922260

1.00000

0.115070

0.171715

0.3612310-01

-0.6966020-01

0.1728610 -0/

-0.100289

-0.5056270-01

- 0.170181

.0-01

0.140835

0.140630

"'-

-0.7990010-03

0.186273

-0.168414

-0.193632

1.00000

yA41A9LF 17NUmArq

In 14 20 21 22 23 24

%)...111c0-11 1.17'15"-11 1.1."613 -o.4 )1/510-11 0./3)4170-01 0.87002)10-01 -0.101630I, nVW.01 1.11"r0-01 0.)111,1l

n.t11179-1.41t94; 0.443244 -0.3918/6

-0.143,696-0.4u/060

1 ^.19s419 0. -0.110'63 :1°3.2= 0.658390 -0.2399660-01 -0.3101690-014 -0.15S1170-)7 -.1.6468150-0.) 0.2499090-01 0.2091143-02 -0.1214120-01 0.4229460-01 0.4449160-01 0.4600060-01S -0.1/21s90-11 -1.6174400-01 -0.1175/1 0.941669 -0.66846/0-01 -0.1509250-01-0.11s738s ..0.ss06.110-01 -0.6206sen-01 -0.16'16010-01 0.7660060-01 -0.137137 1.11=001 0.118235 0.1124027 -m.111,1100-'1 1.711411D-11 -r.144710-01 1:1=9D-01 9.216219 -0.1943460-01 -0.9195610-01 -0.7512390-01R -1.114114 -1.1979,40-91 -0.71704? -0.1198,?0-01 -0.2164179 9.4s4le41-01 9.,e11910 -01 0.140121

0.819669-3.5463630-0i 0.3927920-01 -0.160522

0.600526-0.146773 -0.16/355

11 0.761511 1.501'914-J? 0.0901491-01 -0.550601 0.0446940-01 0.213104 -0.571593 -0.59421211 1.,.1.1044 1.972194o-0i 0.'13327 -0.460559 0.105414 0.410192 -0.415162 -0.45613317 0.1544,1 -1.s4it060-02 0.125115 -0.190,143 0.1841360-01 0.142576 -0.239314 -J.30026911 1.271244

- -0.N151670 -31 0.141617 -0.468112 -0.3310/10-01 0.192044 -0.663254 -0.57140114 -0.414850 0.2151520-91 -0.117699 0.770420 -0.3543760-01 -0.30/453 0.999802 0.9926d715 -0.411155 9.2165710-01 -0.746577 0.771067 -0.4433700-01 -0.104053 0.929711 0.92d/10Is 0.456113 1.1c0458 -0.9430570-01 -0.247921 0.2242280-01 0.206328 -0.170353 -0.157764

. _ ...---.17 1.00000 0.180105 0.466049 -0.474520 -----0.3128840-01 0.713001' --0.435388 -0.43514610 1.00100 0.5074310-01 -0.5107653-01 -0.99do770.-01 0.100752 0.2152d10-01 0.174567D-0119 1.00000 -0.266564 -0.9526231 -0.315350-01 0.303897 -0.323362

CO 29 1.00000 -0.751917-01 -0.395977 0.773384 0.774966

CIO

1P

,I 1.00000 -0.1159031.00000

-0.360306D-01 -0.4423540-012? ..0.308443 -.0.30d29571. 1e00000- 0.9931.7074

9341401c 24

1.00000

'M.-1;A

0q7:5,y)-117 -e.:^7,q,1 -0.,170170 -nt4S 0.1190486 -0.54422S0-917 0.412494

r.?14044"%116169

10 -0.1,444,..0.01,a051-11

12 _1.45711,0 -11)3 -4-.975)&99-11

14 4-.:7019is 0.1N44i9Is -0.1.17e710-0117 -1-.41119s0-01le -0019494,0-9119 .4.4795210-917A 0.:7111421 0.11,14:412 -,4491159-0171 n.121113724 0,117qR5

I.A0ft10

n0ffr

374

Occupation Stratum 12-14 White Males

VA: IALILE

SCift-;01 1

WAGE 2

UNION 3

MIG<50 4

1:).i302670623.4974626C8

0.44J652819J0.41246290E0

STAIWARO OEVIATION

2.3494032601.123903964

0.4901)3412700.492412204

S. 16. 5 0.247/7448(7 0.432040/734

-EXP------ '23.40534451-- 11.60850533--

TRAIN 7 0.3234421365 0.46d1372312

PhY 014 8 2.66468427 0.1-J1663316

NH; WT 9 0.92q7833828 0.914(134749

SVP 10 /.011369436 0.1419310./00

F4IIICC 11 0.667655/1i60-01 0.,!02129t.3110-01-

";',..P"CC- 12 0.4/G49E516:10-01 0.7')93s0(=.4480-01--

Y.:I:ACC 13 0.35J14:66:300-02 0.1:)032215830-01

DA/P'e, 14 SE.44435514 7V.62230237

AW:HPF 15 0.4556345657 0.315816796FrSL 16 0. '[:4337561:771) -01 0.30.;56;04710-01

AlPFAS 11 0.44460320470-01 0.17 891916111) -01

18- 0.'2812893636 0.15/531i:641;-- - ----

rosi6 19 2.53/0';1980 4.50)534737

0NY.:TF 20 0.2430857072 0.3450()33416

".1F-11:CC 21 0.11151335310-01 0.12249)55121) -01

V.ANCC 22 0.771327693130-01 0.93650007480-01

392

00000'T P9..0916'0

4SSZ22'0- 70-01 v4197'0-

00000'1 ISI4C2*0- 70-0.,7,9.27f'0-

00000'7 IO-C.,ZCE09'0- 00000'1

22 IZ OZ 61

'7126.7c*0 L.,Q.,1£0 10-04601S4'0- 20-0002.701'0- 011^0'1 PI

9c7.9170- $70917'0- 99751'3 . 13Z4PT12- 10-0E,OSOZ'C .

0(:'0)0.1 LT

ciA7710- 946771'e- 70-046.A911 Z0-2/01(s0 Iti.pc....v- 1271'0- ..1

f5h4.77'0- IISZTO- 67£119'0 C0-0cc...40- 10-nat544C- 151712D 41

Icqc01'0- Wn9'0'0

70-00./4116'0- 964f09'0

7910I'0 sw.0)16-

10-c4,41c4'0 10-0M4,41'0-

r:f.iLl3- 21:11.t'0

c.ft:;;:: r1

710946'0 040766'0 691.6.74'0- 10-('4144.C'0- 1.49 .,tt'0 1:etql*3- 27

599911'0 t0 -ti1 .7916c0- 10-Off15i£'0 10-01:4501'3 TO-00SsuT62- ZLP,ST.:- TI

SP/112'0- .76c,1'0- 994SE2'0- f0-:11'1',6t,,'0- li91Z1,- 9E1..102'3 01

20-01694,fl0 70-06116,1'0 Z.d...e1'0- Io-t:,,L.711')': 70-0!,1:141.1.- .p411.1e.)- 6 c41c710- .7v/10.70- p4qq.2.7',$ 70-0v1.7N.,:"- 19-C1.el,,,'1%- 10-00.147"1,- P

To-oE5,5pi*0- TO-0eZt.:69'0- lo-egeotst.r., 76-0.5..-1(9'L TC-0/IC(.15t. 1C-Or.i,-aT'2- 1

10-016410.,O 70-r4z047ro in-(f pof.,-n- in-ol.10.t...o- 1J-C444.:1C7'0- 10-0ross.,, 9

2o-asr5e91.0- lo-o....,4.4610- 10-016.7119'0- tlt.6560 ZO-09.21::29'e 119/0.:1'0- s

zo-ocsts.,r'n ozse6r.o- .291671'0-

70-06.7:'0,1'0- 6,7s.,er.c- 10-010.7911*0-

1n- (170.075 0- lo-nssts.to- en-vr:..tvis.c .,te..6/13 12/7C1'0- 10-!:9.14.0- 10-()407,,5"0 1.10,1'0- 11'0::1'r-

10-n=:3- 72.611')

;,,

2

100071'0- 10-0&r.9040- 10-(Ph. /1,..u- 1C-Ci.z.0.-1, 10-11.,7Pt2'1- 10-1!!$:..../.3 1

t13t11:11P1

22 '12 Ve 61 WI AI 4101,1bVA

00000'T ! 91

TO-ORM,O 1.'0000'7 St

20-0569691*0- 196.76.20 00000'1 .,I

70-021:9714'0- I7.2141'0- 70-0100/75*0- 00000'7 C1

9796210- .P74r17.'0- 6f2201'0- /0997L'0 00003'1 Z1

10-02£09/10- 107001'3- 10-0.7f,,92E*0- I.:6,492'0 10-0.62166'0- 00000'1 11

20-nolts.,c0 1','c14213- rotwo- 67c117t0- P5.)671.0- 9.,94c7'0- 01700'7 07

70-n700re1.0- 70- IOtiO7'0- 70-0455/CP0 10-0561016'0 20-0ffl.'21'0- 10-01206/L0- 70k179501'0 cGenI 6

70-0050211'0 tt14c.7'0 4f061'.1'0 FO-CPSsOg9'0 PCW0- 16.7q s, 7701cC- 10-0qE..23 9

10-05.29,44*0- totot)vo lo-ose.,slro 70-019516.70- 70-Ca1I:10V°- 70-0Pc120i., u- 10-0.71111.1'0 lo-rti.vto t

70-07990f7'0 70- ;4+4912'0- 70-n101vsue- 79-(1.7(..,20(3.0 10-01/?.,.,z.0 10-01wtzfEeo 7p-or,114/0.0 z2_nctegzso, o

' 20 -W/45/6.20 /0- 41712"0- 70-0609.'15'0 70-0117112-e 70-M693497'0- 10-066r1o1.0 rc-citt,:(Ae.c- 10-c9/p2c6-3 s _ 10-01:9410I0- TO- !10071'0- Z0-0106/11'0- 10-alst.r610 7o-vver.,:,4'0- 10-04wAt.O lo-v4sI71 .,0- l'-vc6+0,ts.:1- ...

10-o2I1261.0 k7.,f,n0 1/.21o1.0 10-0b29f.6l0- 0711.'2'0- 6,-1,1.4 10-0ov"f_ 14e1,1.7,_ c

10-011O6000 i(-00re, 10-n6976,6.n .117.c1.0- 79-rcq4/sso- :ti/?.0- -..1:,.,I.v Itrf./.;', :1 2

10-0C14?46,'0- 10- 0/1,160 10-nanC66'70 5.6017'0- 10-mTOtl0- ".,0,-..t :)... 47'. r.I 1 1 10-f !.il. i5. *,- I

91

00000'7 10-00S6SWO

10-024 7762'0 10-06cc,0E0 70 20-41700ov1.0 sm.pmro TO

511001'0- 10 691'1'7'0-

P

41

00000.1 9F.701C0- ;7201141'0- I59t.,10- 1/0./420- .c..7010 611541'0

41

00000'1 2o-99420E11..0 10-106F.70/0 Z0-07091P.,C 10-0000604*0- ES.,r1.20-

(I

onrowl 10-01070540- £252.;1'0- 2.274:W0- I/9/11'0-

71

oocovi 10-000.9EtE0

6P9 ri1 '.:,-

70-Vc/940-

11

00000'1 .....

10 -016 / 675'0 10-0LC7W0-

_ . ... 0nn00'1

61C,711(

6

0000..01

v 1

9 s

6 t 7

1

P2m10%

I 9 S r 7 1 41..y 14:6

x1 Alva Ne.111:71i,.:.;:3

376

Occupation Stratum 12-14 Black Males

VARIABLE MEAN STANDARD DEVIATION

SCHOOL 1 9.187500000 3.262166808WAGE 2 2.588046675 0.9999307524UNION 3 0.3671875000 0.4839322812MIG<50 4 0.3593750000 0.4817026070S. 16 5 0.6406250000 .0.4817026070

EXP 6 28.01562500 12.57480569

TRAIN 7 0.1796875000 0.3854355089PHY DA 8 2.6)0625030 0.6668101880

NEG WT 9 1.132812500 0.9338017607SYP 10 6.768281250 0.5215997642tBMOCC 11 0.4792187500D-01 0.432222,i637D-01

YIWFOCC 12 0.9700593750D-01 0.1729455801FOCC 13 0.2215625000D-01 0.5031623421D-01

DA/P4 14 17.25026181 32.05703116

AVEMPF 15 0.3624205729 0.3017795463FDSL 16 0.6339899184D -01 0.1348131672ATPFAS 17 0.361226562-,D-01 0.2092413072D-01

XiININ 18 0.3455067623 0.1842227408EDS16 19 5.492187500 4.962847947

UN an 20 0.1947591146 0.3168350711

%FMOCC 21 0.1192421875 0.2215)18084

%1INOC 22 0.1671640625 0.2345855376

394

CORRELATION RAM/

VARIABLE 1

MUSSER2 3 3 5 6 7 8

1 1.00000 0.330818 0.6577610-01 -0.188533 -0.252421 -0.621415 0.,86112 -0.9547300-01

2 1.3000D 0.395277 -0.253223 -0.200089 -0.219274 0.322695 -0.4756010-01

3 1.00000 -3.6386130-01 -0.7125030-01 -3.7341020-01 0.2341580-01 0.3431410-02

4 1.00000 -0.8377520 -01 0.1466470-01 -0.180904 -0.7277610-01

5 1.00000 0.225820 -0.158374 0.146310

61.00303 -0.265392 0.104441

71.00000 0.4643380-01

81.00000

VARIABLEMUSBER

9 10 11 12 13 14 15 16

1 -0.189179 0.7210330-01 -0.126942 3.7493840-02 -0.5966440-01 0.7008760-01 0.264167 0.5124510 -01

2 -0.105299 0.108140 -0.130760 -0.203510 -0.255547 0.141409 0.500351 0.159135

3 -0.126141 -0.7640150-01 0.9172450-01 -0.186566 ...0.176996 0.3314050-01 0.425693 -0.5223140-01

4 -0.3692470 -01 -0.4391760-01 0.5/41010-02 0.5992790 -01 0.6361350-01 0.7135200-01 -0.2998710-01 0.1190790-01

5 -0.3309540-01 - 0.4823210 -01 0.31E11250-01 -0.7552310-01 -0.6426110-01 0.4495540-01 -0.2418950-01 0.2735390-01

6 0.103089 -0.3453280-02 0.2271830-01 -0.5029130-01 - 0.6064490 -02 0.6243280 -01 -0.137441 -0.3725260-02

7 -0.198090 0.131970 . . 9 8 -0.195126 -0.5452820-01 0.199963 0.132066

8 0.2351290-01 -0.325188 0.137942 -0.8035000-01 0.5895000-01 0.173120 0.7964500-01 0.291808

9 1.00000 -0.169756 0.7536850-01 0.339030 0.369414 -0.4231350-01 -0.254942 0.3759650-01

1011

1.00000 -0.681625 -0.520219 -0.605071 -0.4445450-02 0.1489790-01 0.1019410-01

1.00000 0.195638 0.269986 -0.128173 -0.135897 -0.107903...

12 1.00300 0.955022 -0.178870 -0.345307 ,-0.196933

13 1.00000 0.181918 -0.349505 -0.167377

1415

1.00300 0.545967 -0.9085410-01

16

1.00000 0.120956

VAI'IlBLI 17RINSER

18 19 20 21_,

22__ _1.00000___

1 0.216900 0.4552190-01 0.257853 0.137372 -0.7702240-02 -0.1065170-01

2 0.378133 -0.319145 -0.2463660-01 0.432375 -0.216943 -0.229)40

3 0.4230S0 -0.172243 -0.1682810-01 0.810142 -0.185675 -0.156607

4 -0.5549913 -01 0.110651 -0.161502 -0.3408970-01 0.6124130-01 0.5077290-01

5 -0.126289 -0.132899 . 0.832127 -0.6956590-01 - 0.7356650 -01 - 0.6323250 -01

6 - 0.231511 -0.3937530 -01 -0.117609 -0.107912 -0.4064440-01 -0.3417170-01

7 0.249920 -'.164041 -0.3424940-01 0.140023 -0.202014 - 0.216306

8 -0.234161 0.1062640-01 0.123467 3.8554570-01 -0.4914520-01 -0.2117720-01

9 -0.274914 0.7662840-01 -0.129753 -0.169748 0.348626 0.343368

10 0.163045 -0.304380 -0.3668910-01 - 3.3235950 -01 -0.543620 . -0.638947

11 -0.2565230 -01 0.4129860-01 -0.2852480-01 0.4069270-01 0.214082 0.386111

12 -0.277339 0.625842 -0.2536970-01 -3.261592 0.997729 0.978128

13 -0.291132 , 0.639370 -0.431766D-01 - 0.245333 0.972828 0.968315

14 0.8982680-01 -0.192318 0.111306 0.162705 -0.180981 -0.194505

15 0.505624 -0.312704 0.9579590-01 '0.705686 -0.349004 -0.354578

16 -0.209103 -0.356389 0.9547380-02 -0.2576040 -01 -0.191794 -0.200968

17 1.00000 4.333457 - 0.1697190 -01 0.505677 - 0.282676 -0.271618

18 1.00000 -0.5991410-01 -0.231132 0.627073 0.599708

19 1.03000 0.7731320-02 -0.3075210-01 -0.3429260-01

201.00000 -0.259979 0.237981

211.00000 0.983671

1.00000

0

378

Occupation Stratum 12-14 White Females

VARIAUE

SCHOOL

MEAN

11.426Cd6c6

STANDARJ (AVIATION

1.937899340

WAGL 2 2.354C56522 0.9197505134

UN.IJN 3 0.10434182E1 0.3070491430

t110<50 4 0.5211351304 0.5017133117

S. 16 5 0.20001.00000 0.4011505554

EXP 6 26.19130435 10.74006734

TRAIN 7 C.2C36956522 0.4081548591

PHY DM 8 1.4347E26C9 0.77380.34760

NEG WI 9 0.4173913043 0.7721308940

SVP 10 6.275652174 0.4610421651

%1TMOLC 11 0.139826C8700-01 0.25105036400-01

u&ncc 12 0.7033391304 0.3396325385

%13FOCC 13 0.20704347130-01 0.43677240030-01

CA/P 14 12.60935380 25.23131698

AV8MPF 15 0.29h4168116 0.2595031891

FCSL 16 0.27/J6618456U-01 0.57035768290-01

AlPAS 17 0.1676E6554.5D-01 0.18976085440-01

;;jr%i 18 0.43416c4462 0.2133400297

EDS16 19 2.2CC(C0000 4.569540650

WOW 20 0.51413913040-01 0.1801303133

UMOCC 21 0.73204347E3 0.3370630216

%MINUC 22 0.74602C0870 0.329 ).592867

396

C041-LAIICI% 97441).

vAotALI.Ati"61a

1 1.000007

3

4

5678

2

0.2426631.0 OUZO

3

-0.1666450-020.1701490-021.00000

4

0.4903320-01-0.3397730-01-0.1485440 -011.00000

-0.110416- 0.22J633-3.9955400-01-0.8703080-01

1.00600

6

-0.38361.70.3577880-010.110933

-0.3984060-01-0.2520090-01

1.00000

7

0.7512500-010.2712)93 -01

-0.1052960.6.332360-01

-0.9629110-01-0.131253

1.00000

-41.407302- 0.4034S80.6581210-01

-0.2455940-010.8464990-010.317106

-0.9539730-011.00000

v6RIA61.1hLm,AR

9 10 11 12 13 14 15 16

-(.411.tni -0.3'..5460 -;,.564695 ..47/53(. -0.155014 0.9113300-01 0.0601230-01 0.214081

-C.55t..66 -.).1C 10.1 -0.54315.9 0.316656 -0.445 /55 0.2296b1 0.4C1340 0.161049

r.1I(NII '3.151Gli 0.1521,h0-01 -G./C.3650 -0.7026:50 -01 0.105:32 C.259326 0.16062;33-01

4 L.4410300-01 -.3./132530-01 --0.41(i'.60-01 0.124407 0.4192/0J-01 0.4940260-01 -0.2/06460-01 -0.693762D-01

5 C.2C5257 0.4546420-01 0.116050' -0.105605 0.663080-01 -0.6923960-01 -0.132018 -0.776119D-01

6 0.242039 0.253569 0.271,542 -0.20i.839 0.1492o7 -0.217020 -0.185734 -0.5378190-01

7 -C.114055 -0.11.677 -0.191115 6.233428 -0.116411 0.1737550-01 0.4900620-01 0.4754790-01

A 0.4366124 0.7517:'4 0.095806 -0.539471 0.745056 -0.199511 -0.217771 --0.3785920 -03

9 1.00000 0.2b56(e 0.8121092 -0.565079 0./15079 -0.211550 -0.349416 - 0.8482060-01

1.L000O 0.4755720-v1 -0.650C17 - 0.228273 -0.210308 -0.9827300-02 -0.6612310-01

11 1.00000 -0.431006 0.904065 -0.195804 -0.306211 -0.2111620-01

121.0:,000 -0.122908 0.201566 0.8439200-01 0.376:110-01

131.00000 -0.137366 -0.329888 -0.5925270-01

141.00000 0.527038 -0.7520120-04

151.00000 0.6135510-01

161.00000

VAP1A0LE70,v6CR

17 18 19 20 21 22

C.1 :??4,1 -0.4( esan c.4476501-01 0.4184520-01 0.435067 0.406013

( ..4 :7' .) -0. 14',715 -0.11, 219 0.2640GuJul U.4634(.6 0.235315

C.13''.15 -0.,193.50-01 -C./752400-01 0.637099 -0.219310 -0.217001

4 C.371,570-01 0.11674 4 -0.0035e10-01 0.1100210-01 0.130764 0.130427

5 -C.77758.in-01 .1511"19 G.767112 -0.3611290-01 -0.9600910-01 -0.9122560-01

6 ..i(e..510-01 .4.1 flit -0.1.06360-01 0.664040,9-o2 -0...6046/ - 0.245070

7 C.91 .34290-01 -0 .1/.8135 -0.6!442100-01 -J.4676960-01 0./1d627 0.208845

-1:.394451 4..165410-01 -0.3426520-02 -0.h116/b - 0.352330

9 S i 4 I.) 0.' 1 1 IP.) 1:.1253' :3 -0.7343050-02 -0.57612 -0.425744

IC .) .4. 03141 C..104270-01 0.4666390-U1 -0.496137 -0.901597

11 -C.3 17544 r .o,Stit 3 C.3604550-01 -0.9243330-01 -0.517120 -0.247964

12 C.1 1 si..91 -0.350-7520-01 -0.7843450-01 0.991697 0.983246

13 -L.31C lob .30141.8 C.1533q13-01 -0.14070/ 0.5137010 -02 p.1464800-01

14 C.( ;sc.> I -0./0.12(.7 -0.5489...0-01 0.2694(7 0.245700 0.236153

15 C.3(145 1t, -0.145019 -0.0072670-11 0.425509 0.4728790-01 0.19904710-01

16 C.6064100-01 -0.215945 -0.6911433-31 0.0009010-01 0.3030040-01 0.2934610-01

17 1.00000 -0.381619 -0.4700473-01 0.209609 0.130393 0.104478

le 1.00(1.0 . 0.9(76420-01 -0.151638 -0.517100 -0.497639

19 1.0;..000 -0.7507943-02 -0.3396070-01 -0.3194930-01

7.31.00000 -0.94875331 -01 -0.103951

211.00000 0.997193

221.00000

73- 24

(V

.0-

380

Occupation Stratum 12-14 Cross Race-Sex

VARIABLE

SCHOOL 1

;4AGi 2

UNION 3

MIG<50 4

S 10 5

MEAN

10.342682933.233915610

0.38414634150.42560915610.25497510488

E.FANbARD DEVIATION

2.3499424821.175864652

0.48668962570.49473683460.-4 60586150

EXP 6 28.095121)5 11.46404798RACE 7 0.37904878090 -01 0.1908404674SEX 8 0.14878)4878 0,3560891571

TRAIN 9 3.3024311244 0.4595946639PUY DM 10 2.5)3658537 0.8970097008NEG wr 11 0.8731707317 0,9726477565

12 6.896f:17561 0,4544012312

IM:OCC 13 0.25736985379 -01 0.23041797309-01.A4POC: 14 0.1440500000 0.2730532581)5FOCC 15 0.85317C73170-02 0.27582137860 -01

DA/P4 10 33.82430662 73.'55092587

AVEML)1I 17 0.4275130408 0.3133135309FD 1q. 10 0.4Y48180212D-01 0.86266114170 -01

ATPEAS 19 0.43027926830-01 0.1833778735D-01

OININ 20 0.3079598231 0.1768430866EdS16 21 2.601219512 4.649444776

U'::CiP? 22 0.212265041 0.3323468487

AN:OC:: 23 0.152 5317013 0.2052737215

%MINOC 24 0.1733182927 0.2818013873RAXqEX 0.35365853661)-02 0.92054583150-01

398

.--i 1

oz) cs,

cl

00000't9t

L0-aus6L6'o- 00000"1St

tS9S11.0- S90Z00-0 00000%nt

10-019156Z-0- 06019E-C 8tfLEZ-0- 00000"1It

t0-0£606/00- 9SSS56'C- EtZ/99"C- 10-07P9fL6-0- 03Con't 7t

to-005169c0 et5ELt.0 ttS91Z-0- 00E9E1'0 btSSSt-C OnOK't tt

6E9SSt°0 t0-CP(7.£9V0 SOSCLS"0- LSt90.0-0 SIIS101-0 vSPL6t*0 COOOC't ta

03-009668E-0 016111'0- t0-07059t90- t0-CC0900/"0- t0-CP9t06S-C 10-1t96Pbt-0 to-090/07.9'0 OnCO't t.

LUtZ1"0- 0StL004.0 9ULf8"0 tS/Ltt°0- ISLSRS-0- Z/69S1-0- L6S9S6°6- tO-IsS7796°C- t.

10-09v79850- 0Z£90ZC 10-CIA5ISt°0 11.0661°0 t0-a6tta,t6N., 1,0-(166C446.0 tt97ct-0 IC-cftt64te- /

Z0-0090969°0- L0-3/1E079'0 t0-CIESIZ06"0- 9567c1-0 t13-a987099-0 t0-nct0t0S.0 t0-e0Stc.Fv t;?ttizzon- 9

t0-ccren9ro0 t0-actrs9co t0-USZthSti*O- 519111"0 t0-ctSSf.C7"e- 76061t.0 t0-19t0FL9°C, t0-COE(PLE°0- S

ZO-COLSLZVO- tO-CO9f0LST 10-C969Zi6-0 10-Ct09,,WO t0-04bEi.t9-C- t0-C,.ttcL6-0- te-Gt0Sof'f- ProtrtC- 0

St6PEt"0 9tEtLt"0- LSCREZ-0- P/C7.et°0 07,i-7°1°0 ct000t-0- .171St-f°( TO-CIZEZ'Co'f- f

66atlit'0 ERESZUC- ttat9Z°0- 95tOZ2°0- r9t6L7oc to-crcimsoo to-ncs-407ro rtir,vc z

to-Qom/coo° f111 61'0- tb/SZt°0 96fItSE°0- 10-GGLttLtC r_Stc4:1'0- (S.:JP/VD- 691fSt'c t

e=e6fili

9t St tt ft Zt 11 01 6 21cY'rvik

00000'1P

t0-08z06zh0 00000%L

1.0-Crt0009"0- Z0-OtS609S.0 COCOO't9

10-e66a69t°0- 909Z9t°0 10-000/6Zt "0 00003"1 S

10-09Z9/9/.0 t0-06tLE8Z"0- t0-Gt6VitS.0 10-nS899C5"0- 00000% t

t69StZ"0- 1.0-C9fL9LL'O- t0-01£1912-0 SL6tit.0- Z0-09/t0Q6-0 clonoirt r

ttaSSE°0- tR9PSt"0- t0-OPECS6t"0- fi69fUZ-0- ttSt9tC- L690ft"0 0000C't 7.

10-COL690/.0 ChZ601-0- ttS9E6'C- tZ,W.t.-0- t0-ORLILSCC- 10-J0LRUS'O- 11509r0 000C2't t

npt:rt8 1 9 S ti f Z l SIEllOtA

x1 E111. C01.1V11.1P3

t"4CO

4-0

0U

00000'1 -

ELL966'0 00000'1ZZL6t7Z*0- 0LA6AZ'0- 00000't

10-CZtE00ZO- t0-0A960AZ*0- t0-06SLLZS0- COCOCt

StAOLZ'O859901'0-0t9ZWO-tA9CEZ'O-S7011Zt'O-fES6tS'0CO6606'0EALOW0-C80669'0-ZZ9t01°0-tZtEtS°0-

tALL9ZO 9,11671'04 10-CZOCLOO66SL01'0- ZUZZZO 665E11'0-6970W0- 10-0856011'0 ZO-OL5ZE61-0-96ZLWO- ZEAtZ9'0 t0-0Z6LZW0-tLtOZE0- 9ZSLtC0 to-OC9fIt6A*0

tt6tRA0. 90SZ91'0- t0-019191E'0SL0966'0 OZZAAZ0- t0-CFP9C9Z0-99106I.0- t0-0Z6tSACe 10-0t8ArCS090089'0- t0-OSLC060- ZO-040CL90-AZA0610- t0-00ALL06'0- t0-crtontfoIZSEA6.0- 6871(6'0 t0-01.10SrZ0

00000'1P69ALt'rSCO9W0

t0-OtS3C65'6-SCt?(I'C?EtZit0ZC9611.'u-

Eu(91.'0-10-0169;01"C-

erfisr-c

L0S0(7'0rfr4Pt't.-

SSSL01'0IC-rOSL(1,_0tf-cnr6cto.c_

prtt;?'0Sit80t'0

t0-c97t71Z0917ezt'c

10-CL1PSsS-Ct0 -C[687106-

StC4S1-0-t0-r?Pci.1%)-

SZ

SC.67

tZCCt?OZ61EtLl9tSnt

itZttt

Ot6

L

:11r V 1 gVA

001.00%LISLWO- 00000'1c975ZC- 81,091t0- u9000'16LOSE1.0- SC61a'0 t0-OZCL6AS0 00000'1

9SSL6tt- 0grat0- 70-0091801J Stat6o,0tzoc610 940z67.0- t0-aubst,60- At9ASZ0-

75E6C7'0 CASSAt0- 6cillOt0- 9h7LC7'0-

ZO-05COA9C- ZuNLOZO- ta-UCICeZto- PoLLQ-Z91L61.0- A0A1i.ZO 10-001LtrAe te-CuFCL6A'0-

to-ast ci Lr 'c oqz,.Ptzo- 7o-aisckni)- t(-CrSALZUC-Et L8Ct '0- t0-0Z8LL9t '0- ?At( (1 '0 Lt.9L')Z '0

t0-03A07.A9L'0- L0-006(9600- t0-0t685c.S0 10-0167971'0- to-0z4sbcz'r- 1C-0n6r1rZ0 t0-OCS91/10- PECSIt'C

Le6(0'0 ttItAP0 50700Z0 - CO-M7(00- 0o7L 7( "0 77S1,9t0- tu-n?(Asto- SSIC6t'°-

t0-09t6tSS'0 t0-0A60L010 t0-Cfm/Ait'0- Ct6Cit'0 9(L77tC 10-06161L0'0- ZO-C87Lt16''; to-atepf.L4'u-

t0-at.zzt9C0- I0-ontothuo- t0-a8 St0It'0- t0-CN1f6A0- t0-CrociAtC- 10-OtilCo,.:U to-Ott0:C.:e to-rfrItn-

t0-0AOSLSZ0- L0-091909E-0- 10-00R6SA9'0- 1,91.L66*0 tO-Ict9Cc,h0 7c(;At'0- Zo-C(AtA01- IC-CSAL916'0-

10-00E16;6'0 10-O9tfOC6'0 10-cA0(6660- t0-00(t6P90- 13-Co617SC LO-gto7rOf°- t0-01r0iiti- t0-0SZAtu8"(7-

9LPSPZ'O- t8ZZ6C0- LSZ60t 0 95t6t t '0- LI gt (A 'C- C6t1140 tO -rin. 9 tAF'( CSSStr 'e

t ASSOC '0- IP( APC 'o- 1S8SE t 'e c.OF Act '0- (4'7' , Li 'C- Lo tf,4,7'1, it .; Ll't, CI. It.t ' r

t0-C9ZS6LL '0 68SSOt '0 tO-of t (.98s '0- t0-Ont i Opt '0 10-Oro. ARPA '0- tOuvtu 70.-CA1t6'0- ti-q.bitts(%

AZ CZ Z? t Z OZ 61

67.

ZZl4C'61atLt91StAt

Crtt01

L

S

42;f1.1r9

$f Lt 37rY:-:YA

383

Occupational Stratum 15-17 White Males

VARIABLE

SCHOOL .1

MEAN.

14.21254355

STANDARD DEVIATION.

2.656956919WAGE 2 4.831254355 2.038005974UNION 3 0.1114982578 0.3152979263MIG<50 4 0.2682926829 0.4438647069S. 16 5 0.1846689855 0.3887065855EMI 6 22.43902439 10.60325239TRAIN 7 0.4494773519 0.4983098062FIN OM 8 1.5365E5366 0.4995307366NEG WT 9 0.13937282230 -01 0.1666646362SVP 10 7.477700348 0.4333851233tomocc 11 0.93170731710-02 0.75043452080-02T.WFOCC 12 0.96745544600-01 0.1127932453TBFOCC 13 0.28710801390-02 0.84556080540-02DA/PW 14 50.22205967 108.2667440AVEMPF 15 0.5541205575 0.2794237727FDSL 16 0.45198488C70 -01 0.731/900541D-01ATPFAS 17 .0.47249128920-01 0.18929254630-01VIININ 18 0.2821773328 0.1419488676Ens 16 19 2.666411150 5.743217312UNXMPF 20 0.64912543550-01 0.2064812091TEMOCC 21 0.99616.7'4740 -01 0.1184538186%MINOC 22 0.1089337979 0.1210161421

..

401

CORkELATIC4 4ATRIX

vAtatA0LE

12

34

56

78

NU

mbE

R

11.00000

0.433729

-0.224386

-0.6330120-01

0.599980D-01 -0.290918

-0.8554820-01 -0.2035330-01

21.00000

-0.163786

-0.137325

-0.4707170-02

0.4498990-01 -0.7960690-C1 -0.136739

31.03000

0.110300

-0.61C0330-01

0.6688410-01

0.1312430-01

0.107209

41.00000

-0.105782

-0.5260500-01 -0.2544860-01

0.2654033 -01

S1.00000

0.9104200-02 -0.3289520-01

0.4611640-01

61.00000

-0.189681

-0.1961810-01

71.00000

0.5310310-01

81.00000

VAR1481E

910

11

12

13

15

15

16

N048E8

10.5641240-01

3.357VC8

-0.115162

-0.220490

-0.4512110-01

0.27/9450 -01

0.8641470-01 -0.4309770-01

20.4050680-31

3.289109

-0.205257

-0.215080

-0.104085

0.16984,0-01

0.117782

0./932270-02

3-0.2967553-01 -3.5082860-01

0.120959

0.6254350-01

0.1590250-01

0.4761550-01

0.34818/0-01 -0.2900470-01

ItiPh

4-C.53/25/0-01

3.1848760-01

0.43/0150-01 -0.1888043-01 -0.513J95U-01

0.4106460-02

0.4249940-01

0.590518U-02

95

0.5807623-01 -3.2320740-01

0.5057830-01

0.150313J-01 -0.25/0950-01

0.20040

0.3105600-01 -0.2628600-01

6-0.5887450-01 -0.18/1040-01 -0.1315633-01

0.114797

-0.2003530-01

0.3436213-01

0.4353003-01 -0.2001150-01

7-0.7569360-01 -0.1299C2

0.7021/90-01 -0.10687/

-0.5392620-01 -0.6796930-02

0.125160;

0.5166853-01

80.7784990-01 -0.228792

0.438,45

-0.125090

0.114943

-0.6407800-01

0.2587801 -01 -0.1597790-01

91.00000

0.101134

-0.7064050-01 -0.7048990-01 -0.2849180-01 -0.2065570-01

-0.3309233-01 -0.5103080-01

10

1.00000

- 0.391196

-0.186984

-0.152783

0.159791

0.7441083 -01

0.2742260-01

11

1.00000

0.303501

0.346529

- 0.6740370 -01 -0.8175000-01 -0.6839390-01

12

1.00000

0.648161

-0.5568010-02 -0.119512

-0.5684820-01

13

1.00000

-0.2909530-01 -0.149540-01 -0.5711630-01

14

1.00000

0.445996'

0.6626760-01

15

1.00000

-0.2166510-01

1.00000

VARIA8LF

17

18

19

20

21

22

N04318

10.8631130-01

0.1968750-01

0.175272

- 0.173357

-0.213174

-0.215810

70.160617

-0.7501250-01

0.1924460-01

-0.133110

-0.212232

-0.220463

30.15)1820-01

-0.4252470-01

-0.9940710-01

0.888099

0.6068990-C1

0.0690460-01

4-C.190819

0.3434600-01

-0.919270-01

0.105283

-0.2164650-01

-0.1815540-01

5-0.8198660-01

0.2969220-02

0.984556

-0.3650250-01

0.1241180-01

0.1534970-01

6- 0.129963

0.4/27710 -01

-0.1102320-01

0.5850420-01

0.107831

0.104766

7C.2641793-01

-3.119888

-0.1854190-01

0.3427290-01

-0.131309

-0.100681

8-0.9427553-01

0.114186

0.1)35550-01

0.153858

-0.1/0913.

-0.8116830-01

9-0.7,640^10-01

-0.9358000-01

0.7701530-01

-0.2033140-01

-0.:,415530-01

-0.7207040 -01

10

-0.1163470-01

-3.1(4980

0.1853300-01

-0.6213303-01

-0.183954

-0.209209

11

-0.6471910-01

0.201534

0.5017963-01

0.7442690-01

0.313734

0.369097

12

-0.3884180-01

3.131022

-0.670120-03

-0.4200900-01

0.598523

0.996185

130.7204340-01

3.109916

-0.3121920-01

-0.112131a-01

0.689142

0.696028

14

-0.232797

-0.26088

0.197531

0.130150

-0.7318350-02

-0.1140220-01

15

-C./844520-01

-0.226171

0.7849130-0(

0.165452

-0.114868

-0.117504

16

-0.140003

-0.240545

-0.1555610-01

-0.2510863-01

-0.5820970-01

-0.6121040-01

17

L.00000

-3.175142

-0.8216560-01

0.8600960-02

-0.3184100-01

-0.3518090-01

18

1.00000

-0.2775160-02

- 0.9376420 -01

0.112607

0.142790

19

1.00000

-0.5705070-01

-0.2814020-02

0.3523560-03

20

1.00000

-0.4080200-01

-0.3501210-01

21

1.00000

0.998265

22

1.00000

385

Occupation Stratum 15-17 White Females

A r

c." 4-Irtt

WV;F1,M1'1%)

4!",eqlC. 14r-vt,

T4A11iv,

7

1

4

6

7

0

1FAN

12.966666672.831000000

0.10000000000.33333313330.16666666672.5.71311333

0.211000e00001.466666667

STAN0AD DEVIATION

2.2815105431.063703193

0.30512157660.47946330150.37904902189.818818432

0.40681810220.5074162634

Mr^. wr 0.010 7.513133313 0.4256715326

1!e: 0.90313333330-02 0.82314926510-0217 0.3241666667 0.2454937998

.vorn7r 13 0.1106666667D-01 0.17101085530-01r1 / n 4 14 .24.22796968 32.88062392Av-.4nr IR 0.4112411111 0.2803001130

r.r.r1 16 0.28406134540-01 0.41191642730-01ATnrAs 17 0.48143333310 -01 0.18323143210-01Y111 %!! 14 0.1014732116 0.1359390593

10 2.200000010 5.067815962ivoyqpc 0.79844444440-01 0.2499296122

71 0.3353333133 0.2611869899TMtmlr 0.3443666(467 0.2597110441

403

38-6

Occupation Stratum 1C-17 White Females

4 .4 nSO 0 0

1

O C3 CD.3 4. .3 .1 7 - 0

m ) 1-en 033 7 4. .3 4. Ws CO JO /-0 r. - N 1.1

01 r, Int 41.4 fy md ns

0000000

Ct 00, P. 0l0.0 0 0 0

r, 0 .0 P. ^4 CP C)CY 04 4 r4

C) C) 0 C) 0 0I I 1 I

N . nC) C) 0 C) C) CI 0

IIIC) COO 0 0 C)

7 4 13 13 7 -. en en 4. n4 "0 PI .4 0 0

C) ^4 CD r. 'no r- tn 41 C)

4. 4) J.r on Pn n- .4 on y. .2 C. .4 .3 0m 41 el -. nD ry 0 .- CP .4 II

0 .o004100..0.411, a .40000000000000001

.11 . 00 oI I

e- C> -J .0 7 ->O r4 rn C) 4 .4 CIO m 0 4. Cl .3 7 m 4, 4 U1 C)Cy 4, CO CJ CP 2. W, CD C) C1 a. C)p, ys f 0 .4 el ry 1. 0 .* 4. 0yrs 4, ,c) C) cy 41 .or S uft cc UN 0-4 r4 0 0 r4 el 4 4).. .4O CD C) C) C) C) C) 0 0 C) C) 0 0

.4 w48. 04 04 .4

0 0 0 0 0 0 0I 1 1 1 1 1 1

0 0 C) n C3 nP. c 4. 1 CO CP 14 C. 22 0 C) 0 0 P1 0' 4) .., eq 0 ,, ..., us r .4 .4 .4 w4 4) P. rj On 4 ../. on N P. UN .0 N04 , 4. to e n m 0 P. . ...- .1. P ' 0 = es s; ...1 4. 0 Cr r 1/ P. C) .1. sr 0 .. 4^ a 0 ... P4 C. in .4 0

.. .4) ..,, N 4 3 u s 0 ..r r 4 T A ) ni - e 0 C C .. 0 4 ) r - 0 e 4 0 .0 C. .0 V. 4- an 0 C. a .3 14 Ca 0 ...4 1.4 C el UN UN C)

.. r ) .4 el 0 C) ... 7 n4 13 rw - 4 41 c , ..? 0. .{.. C C 4. 0 PNI on ' C O ' ' 0 41 7) CY 0 1. 041 C) C) 4 CP 0

.... 4, C, 4N 0 ...) t- ,, . e - 4 4 4 r4 ty Co 0 02 a' P. x ...r J-. a) ....1 0 (1. ,... Z! MC -4 en r- .r POen.. .4 .4) ri

T- ... .4 ey CY n+1 CD 4 4) .. N 7 P.. w4 ry ... .4 oy 4, 0n -. C C ,. 0, ..- o, ,.. .. ,.. C'

. . ..4 O .... 04

0 0000 0 0 0 0 0 0 0 0 0 0 0 0 0 0000000000000000000001 1 I 1

1 1 I I 1 1 1 1 I

... ..4 ...4 .... .. .. ... ... .. ..4

0 0 C 0 0 0 0 0 0 C

i i I i i i i I i I

0 0 C 0 0 C) 0 C5 or s n4 .4 .4 C I nl .y ry en 41 C . C Z en 0 C r 0 1 0 NS 41 ga Cr ar P 0 NO .0 C) -. C .0 ../. .3 C) oo JOt. P. f es 0 r, w, 1) 4) 11, st 0. CY CC 1.. P Ca

e. ,e 0 sp ,f 11 0 .0, Us ,r 4. e. r- 0 011.1, ... 0 0

nr -. 71 c..... ../. o . ..0 0 ...c ..? ?...e 0re.4 ...r es, .40 . on .4 V. 44. lr T, 4) 4. CD ,.1 .3 co en. .4. ...2 I- r. en CO 0

w n 0 C P 4 ; 0 4. C) - 0 -. .3. .3 ...rry .4 CSO C) CY er 4 C. 4 4 c 4 -, 4 ',0 rj -... . . .0 I 1 - k t s .4; 0

4, .1. 4- C1 0 .. ..: .0. CC 4) n .0 0 1.- .. 0 1. e- r. .... PJ P. -. .0 0 CP C .... NT C. '., C.l rsi IA 4 0

.1. c .4 .4 ce ... -+ rn .. ..4 ev 0 C) CI .3 Cr3. 43 ... r.... .... t. m Q v. .... C. C. .... en in. en .4.4 .4

. . ..4.4

C 0 0 0 0 C 0 C 0 0 0 0 0 C) 0 0 0 0 0 C Ca 0 0 Ca C) Ca Ca C) C) Ca C) C) C) C) Ca C)

i i II I 1 '

I I i 1 1 I

4:1..,, ... ... .4 o4 .4

..4 00 0 0 0 0 0 0I

1 I I I I i I i

CDre m c, ,:, f., 4 r C . ..,

J cc r. 0 7 ,0 .. na to t- cC f- .3 0 C) cr en r- en e- le 11 es r, 4 4 c1 CP Ct ..r 1, 1.- CD

43 23 C' 0 ,r 45 4. C) e. 03 us 4. .. 4, 0 .. '- .0 0 ..4 .2 0 V. 4 ce CD 0 4. r. .3 rs 1. 4) C>

,r 0.3 .on 0 ry r. .- 40 r4 0. or ev r) r, 4 C) C) C., e- tr en 7 ..r 4. ,.... 1- -. 7 e- C .3 C -. .r0... en ...r 0 .. 4) I' x - -. f. CC.. Ltn --. 0 INI CC 4. Iv 0 2n a*, sr 1-2 u- en 0 .2 .1 a c o4' 4 CP 0 r4 c N 4. I. i "r 0Y+ 0. ...... 0 4. e "- 0* .: .: .1" 0 7 ../..... - Cr 0 0 Ic 1.- 4 CCV Psi eV

0 ..::: .4 I", .4 4.4 41 0 .. N .. .4 7 L. .3 .4 .O ..4 .4 .4 v. ... .. 41 N w4 .4

.4000000041e.. .. OoOOOOO. 41*. .4

CD el 0 C) 0 Cl 0 0 0 C) 0 c) T.0 4-1 C) C. C C C: C, C, C. C C CI 0 0 C) 0 C: 0

I II 1 I 1

I 1 1 1 1 I I I

p.... ... ... ... ... 04 G.

0 0 0 0 0 0 01

1 1 I 1 1 1 1

o a n a a o a ac t. 0 It 7 e. 7. 0. tv en C 0. 0 r .. 7. a: .. f- 0 u- cr C ee u- m C.' 11` a a4 r. o r, 0 4 V' e C. es 00 C.0 C' C f- 7 C "*. .1" r 4 of a en e- ...1 a- 0. 0- 0

e. 4 C .4 .4 ... C. .4 0 .1. C C) CP C a ..., P. ... 0.. 0° C. ".". .1. r- -- .. n. en eno 0

M .. 0 .4 C 7 4 ev C)... v. .-. 1.. ... P.. N c t ... P.- Cs N.. ... Is (1. 0 .0 0

.r 1.- 0 nn ... 7. en .. .3 m ,r -. 0 r- .C. .1 we a* .0 es. r. .0 .1 kr -. - CC - 0: kis 0.... ... .... r "" 0' .0 Y- r 0 4' -.

CO.4..

..

CO CCCCCCCCCO CCCCCOOCC OCCCOCCOO1

1 1 I 1 1 1 1 I I I I

.4 .4 .4 .... e., .4 IV a.. 04

C 0 0 0 0 0 n0 oI 1

i 1 1 I I 1 I

Mo ...-..c C C cc 0

CC rs C r sr V S 0. en c 0 0 r n .- C.1 - 4. r n cC C.. .0 0° 0... NV 4 - 0

st, C C 4 - C C 4 0 - 0 u- e. ..0 C n. - C te C - ..: .1 N.: .410 0N 7 C ...et's.... rl ". 0 a' F .. ..: 1.,7 .1:-v c' :'.: P: ;:: fr ..7. f-::- 'C :!: j. E:

.-. 0 C 0.- C C ... tt 04 a' C. -.0 C. -. c ,:, .. en .. cc ..1 0 C -- tr .- ..4 f. 4. C 1\ ,... ". r. -.. r II 4, C)

V.(... .. ... ..,C ... .- fr 11, 0 . e c.... .r -- Cr .04 .4 r ,.. o .... v. .... er,

... . . . . . . -. ............ ....C rereCr,reC e c` r r r e. e. r. r C. r r corer

I 1 1 1 I I 1 I 1111 II... rn ..., .. .-4- r C .- C.

I o 1 1 1

C C. C C C.,r.- .- o cc 7 r e. r. C u- u- C 7 r- 0. x . .

p c. . C . o e .e e r. I V r . U 4 ,r 1( - c CC r)

Vc e 0 I. V t .4 P. 4. 4 c tc-ccor-a

41 C.... P. P. C ...0 C C 0. f. rCCCee t 0

a r c c C C 4. .. - 0 0 k? .- 0 r Cr r r r r. r r. C C -r r r r rver..)rr4r,r

: ... .-

C cc,..c.cctcc C.CCLC-Ce CCCCCC.C.CC1 1 i I

...

... 6 4 6a .. .0 -cr-1 .. . c n .. - r,, .. 'I .. c a r - e - . v . 1. a C r .- e e ,f4 r . ,11..P.c c - e. . 4 Y. r a reef. a. ... :

gm ..... .. r.

... 7 . . >

404

vAQTAPtF

387

Occupation Stratum 15-17 Cross Race-Sex

MEAN STANDARD DEVIATION

SCHnOL 1 14.09006211 2.639496435

WAilF .7 4.628322981 2.043987532

UNinm 3 0.1149068323 0.3194058308

MIr<50 4 0.2763975155 0.4479118082

S 16 5 0.1801242236 0.3848890079FYI, 6 22.71118012 10.51451152

QACF 7 0.15527950310-01 0.1238323690

SrY A 0.c9378881990-01 0.2996356863

TnA1M 9 0.4254658385 0.4951829648owl nm 10 1.531055901 0.4998113071

NFr: WT 11 0.12422360250-01 0.1573763725

SVP 12 7.476260870 0.4322217690

T.P"nCC 13 0.93229813660-02 0.75398029870 -02

crvFnrC 14 0.1199813665 0.1461922292

714-cCC 1' 0.3613540373D-02 0.97812137570-02

FIA/TP4 15 47.30596888 103.0200964

AVF:inF 17 0.5407351967 0.7819909417FOCI, 18 .0.41449043370-01 0.7050367285D-01

ATPFAS 10 0.47670807450-01 0.18512469960-01?up1IN 20 (1.28355C2783 0.1404519636

Frci 71 7.599379882 5.642329968

iroXYQF 77 0.6964068323D-01 0.2153510186

Tr"nrC 23 0.12261490f8 0.1515361206

7tArl' C 24 0.1319378882 0.1550108102

RAxSPX 25 0.62111801240-02 0.7868818624D-01

405

388

Occupation Stratum 15-17 Cross Race-Sex

A A A .400000 0

P. N 7 ca POM 1 fa f caCO M PA 01 el ly .0

.4 A .4) 4 1 .0 0-r 7 0 0 0 A N 0- kr) n1 9 Aite,e4o0 0 0 0 0 0 0

I I

.4 4 A 4 .400 00011 111C3 C3 0000' rn PJ JOey 0. P. C) .0 C)

P- r4 to or 44 .o CDO A y P- C) 0 001rn....., CO 0001C) CD 0 CD CD 0I I I 1

fy .4 A AC) C3 C)11110 CD 00

*N1 .0 .1:2O 1, C3 JP 0- 1.nI 0 0 0

ci .r0ftf0 .40 0 0 0 0

I I

NA000I 1

0 0V CD o% C)

01 0. 01 d 0O U oN C 0 0

0 A N C4 .4 C .4.4

0 0 0I

OO co P 0.OT 0 0. 04) Pro% C)CT V% O C)ec .4 4.

.4000I

4 CC0c 04 VA CP-Ctf 0

ry w40 .40 0

I

.4 .4 ey A R A Pv .4 A A o0000 C) C) CD C) 0001111 1111( 1 11 10000 0 0 0 0 CD 0 C3 0 C30" 4) 4 n't .4. J OJ oy o cp C) nn 0P- CD kn .4 en en kn 4. 4, el .4 .0 V. 03 01 C3

4 O e. rn 4 0 Tr. A yr g) ry 41. 4) 0MC A Cr. 44 a. .1 4 f.1J ry TNNNO(7. .4 4 P. OP 41 P. 4) .0 MJ orrn J yr ry ry .4 en P. NJ A A V4 1.1

4.0 A

C) 0 0 0 C) 0 C) 0 0 C) 0 C) 0 0 011 11 111

.4 ey .411J A 1,1CD CD CD CD C3 C31 11111O 0 0 CD CD 0un CD P. CT .4 4p .4 4) 41. cp p Of A0

NN yl p. .4 U1 .5 .4 on f4 C) NOM crrnen ACM A A 4100. -1 0J1 010

M4 . p, pm 55 OM J0 ,r 4) 43 C341 01 .r o4 0. 0 1 CD 4 CP CD P. rn 0MAN fY y1 A fn fY .4.4 N ry ey p A0000 0 0 000000 00

1 1 1 1 1 1

4 P./ 4 .4 e4 eo0 0 0 o 0 01 1 1 1 1 I0 0 0 0 0 00 d0 4 P N0 N 0, 7 17 N 0

4. P. ey pft CD r4 .I. cm Pen 03 0 on 04 .. AN an .0 4 P4 rm .0 CD an .4 ei 004 P.. Cr. P. 7 ry 4) c3 r, 4. 44 on Cr 74 C37- NM C) .7 P. 01 P. 4) 01 al d0

A PY A P. f4 .4 ff) 4 A 41 4 AA .mef00000f0f00C C ) D C D C D D D D D D

A1 1 111/A A A A A MI A0 0 0 0 C 0 0 0I 111110 0

0 0 0 0 0 0 0 07 rN , ry 04 A .0 Pm Pm Y1 ,r 0 0.0 c", o" C. 7 P- C 0 Nn . J0

0.% of C. 0 4' - 0 r -.....r o -. o... 1 4 .0 AN CP 0

4 ) P r A J PO 4 1 . 4 4 73 c0 41 y5 CP .77 AA 41 CO 4 A P. ...1 4r 4) en

ow.o 0 o 0 o 0 o 0 o o 0 o

I I

0 0 0 0 0 0 0I 1 1 I 1 I .0000(3C) Cc) y 4.1 A 4) 4) CO 0 CP C)

C r+rA rnNN P. 0 4. 4. ct) 0N. 4 09 NNP0 P 1. 0

C 0- O. co *2 u3 O 00 4 0NM CD .4 CS y) A en CD 0 411 0

f, ANN u. N P4 CP

0 0 0 0 0 0 0 0 0 0 0I 1 I I

.AO C 0 0 0 0 0 0 0 01111111111O 007000000

LP A 4 4 r. o. N 0ce. ey o

C- 4) CD 4) CD ,C PM CP oNC 4 ca c3 0 NN ON 04.4 z, en 1, C Ce rbO7u JN 4 .0 4 7 e. e.C) C C) C) 0 0 C) o 0 C)

11 011144C 7000001 1 1 1 1 1 1

C COOC)00JO ..r CP ea cr 4m 4 CP 01 MJ C)

CO 00 or .1 PY CO 0 CT, 01 4) C)

ts, N 0 C cr cC 7 1, 0 .01.. 0 .... .4 ,,, C. cc 0, if, 0/ JO4 0 v.: C. e. cC .. rn p- 0 01C.4' N. -.. - 0 1111 1.. ... N ...,. A 0 410.00f00cree-0c+oe-e-

I s

.. -. N ... -. Nr C r C. C etIIIII 1

M C C. C C C... C .0 fs. ct 0) 0. a 0 ocr e- ...,, ,:. c c .0 o P. P cl... A C *7 0 c .. 4 4 4 ct Ce C

C 0 .1. t. r. e. C .0 p. C2 C. 4 r c c r, o 4 o, " v , 4 N 4 -7 .... ilgfee0f0 ACm

CCCCO 0 O CA 1 1 1 1 1 1 1

om U. tQ -.. 0 .J CI... C Ly. 01 5 ... r. r . 4 y 4 . az <6 ... n r 4 it 4 e. a 0 c - ev f' 4 V 4.7 ... ''.

m. SCc .7 a Z'C

a< 7 <4- 7 7

406

$;11/

414

IIA

RL

FN

1148

FP

4 7 P

10

11

12

11

14 lc

1,,

17

10 19

1P

71 27

2-A

74

WigtARts

olgunrg

11

(1.1c/S4c

.*AN-r*01-01

!`'27[.19 -01

0.4121710-01

0.12,.0410-011

-0.11"",c10-11

t.114744

0a16A10-01

-9.7757400-91

A.15c4A70 -.11

-0.4117370 -01

- '1.106316

9.7795140-01

0.444ns

1.90090

's

In

-,.2071540-01

).4271150-01

-1.154(!910-M

-1.3062093 -07

-0.2144:40-01

-1.1117240-01

-1.1:124081-01

-n.711co00-01

0.1-0710,11-01

0.1717810-02

-0.4979480-91

0.2099390-01

-9.50.30340 -01

-0.6912219-01

-1.4410700-01

9.6862840-01

-0.9156410-03

1.00000

19

0.7333120-t61

0.145.,73

0.7;16402D-01

-0.P475100-01

-9.1 (1009

-0.11217e40-0!

0.116658

0.7607190-01

0.7919140-02

-0.6osiAlo-0!

-0.2682190-01

-0.5318100-01

-0.6097640-01

0.4015450 -01

0.4876410-01

-0.272184

0.1232600-01

-0.117962

1.00000

20

0.114417,3-01

-0.014n600-01

-0.2516,160-01

0.7060 060-01

-0.1014110-01

0.%06,,010-01

-0.251(000-01

0.4100769-01

-0.121!79

0.112247

-0.96660-01

-0.9138983-01

0.1331571

0.8635340-01

0.7597600-01

-0.261290

-0.202611

-0.234239

-0.192116

1.00000

21

0.125d45

0.2',09470-01

-0.105150

,-

-0.10071

0.904401

-0.15930-01

-0.5194860-01

-0.1165790-01

- 0.4554770 -01

0.469,510 -01

0.7574810-01

0.7914180-02

0.8123550 -01

0.6464180 -02

0.5490470-03

0.209211

0.1795800-01

-0.2964040-01

-0.104993

-0.1432520-01

1.00000

22

-0.1844S9

-0.113316

0.898904

0.9676290-01

-0.5399790-01

0.5095470-01

0.122761

0.3318020-01

0.1057310-01

0.160413

-0.2560540-01

-0.9321260 -01

0.8751840-01

-0.4957110-01

-0.2458400-01

0.11716/

0.22175$

-0.8830910-02

0.5307570-01

-0.6417480-01

-0.7073890-01

1.00000

23

-0.191309

-0.2649e5

0.1164030-01

0.5810700-02

0.10.1982'339-01

0.140104

0.3587120-01

0.427317

-0.172965

-0.4462060-01

-0.6220220-01

-0.193e01

0.171973

0.99q070

0.764713

-0.3237900-0r

-0.100269

-0.6802920-01

0.4452570-01

0.8706310-01

0.6189970-02

-0.4876620-01

1.00000

24

-0.194173

- 0.210913

0.1769130 -01

0.8593550 -02

0.2352630-01

0.116460

0.3721540-01

0.423641

-0.168002

-0.2209020-01

-0.6484030-01

-0.213718

0.21(977

0.997457

0.770799

-0.3422230-01

-0.101118

-0.7064550-01

0.4113620-01

0.9506660-01

0.1017970-01

-0.4404530-01

0.998851

1.00000

I-n.176q980-11

7-0.740'4110-91

.9564o711-11

40.1452760-Q1

-n.17(1654E1-01

61.3171750 -01

7c1.cl94R4

0.717943

0.11",P30 -31

In

0.7624 n,0 -11

11

-9.C75 ^9 m1-17

12

-n.47,4419-11

1114

_0.1A4p.,,541 -:/i

14

..(1.141,11,,2n-37

II,

-0.7403170 -01

17

-0.4S46c911-11

IR

-o.,,N6p^541 -nl

14

41.?306290 -01

7n

0.1 22 16141 -)1

71

-n.1647760-11

2?

r.7007070-11

-0.1819910-91

24

-9.1427050-91

2%

1.09900

cn 0 In 73 w X

390

All Occupation Strata White Males

VAPAAULT:

1

MFAN

10.94864'165

SrANDAn oiNIATION

3.0 34119469WAGE 1.42 3854054 to.? s 14') I

ON 9EM 3 (1.,,186486486 9.491717'1177

'I G< 4 0.4061081081 ..r9 16 1: 140

S.AT1h 5 0.?114524374 . 445331564 ')

EX0 S 17.74176173 17.15366031TRAIN 7 O.Y5/Z91273 3. 4372(sz79 4

PITY OM ?..36P1a6108 9403$3510925

WT 0.-;4541154054 1.904494530110 3.'61351351 02.5699,492

Svp It 5.,/6,q/595 1 . 995143903Tilly:CC 12 0.4.1=)342142.20-01 ). 52973345160~01r:10:1CC 13 1.12725021o3 1.1,483h4it105ne.:1F0CC 14 r.112/02C0903D-)1 '1.:'17543.311U-01DA/PW 15 14.9.36C7,90 :10.";56?7515

NYL/PW 16 35)753 11[,5') 5. 4 )6 )3480AVf7"PF 17 0.4:'1.-%10702/ 0,:qq78%/4472

11 W.1;1462196!)-(11 0.:.7)041'..9310-01

FrISIA:x 19 C./2217117410-ot n.q?)to114??0-01ATPEAS 21 r./.,.hv.21A2M-)1 1.1755434:50n-01MININO 21 0.2ri/775.62 0.15M1/5216EnSib 22 2.7/045)45 4.=055066517

HNXvPF 23 0.'00,15;1792'1 1.19717509211-mn0C 24 ^.17149C2703 ).19'6212790

PIN1CC 75 0.22/0264805 0.1971R35872G9 +SV 26 3.P27945946 2.7156U042GOXSV 27 2J,7L6C4865 11.00030345

408

C.1;

CP*Z4r1ATICk 4ATRIX

W.,41411 E 1 2 3 4 5 6 7 Cl

1.10130 1.412129 -0.218874 -0.135591 -0.150448 -0.474197 0.225483 - 0.333458

1.00000 -0.4900770-01 -0.153548 -0.112269 -0.111559 0.106515 -0.2;0593

1.00000 0.103967 -0.106101 0.641,5140-01 -0.8913650-01 0.333232

4 1.00000 -0.6345390-01 0.5991220-31 -0.83817/J-01 0.7006500-01

5 1.00000 0.2903040-01 -0.5465310-01 0.8508790-01

6 1.00000 -0.418:014 0.104261

7 1.00000 -0.6443150-01

81.001100

WV,IARLFr,"r6

3

4

5

73

In111211141516

-('.1.q/1-p.47cvl?9-11r.Q/14471-)10.371'171-020.9145739-010.40P8440 -31

-C.1724770-010.4891681.00000

10

).4-%-(015

).11t.0-1.4.1..704

-1.1,1518-1.5/10143-1.1473090.216918

-1.365029-1.145601

1.00000

-01

11

0.1 .l.?1430. s I hlt. 1

-0.1',5421-0.12,.071-0..2:340.:10-01-0.1072420.223500

- C.9.50730 -010.1413730.8492441.00000

12

-0.191421-0.24'.so50.2.194640.1109430.4444670-010.119112-0.1887550.4520420.164491

-0.654721-0.605660

1.00000

13

0.6231160-01-0.81,1,7) -01-0.9244490-010.5J75510-02-0.4581650-01- 0. '120n/50 -01

-0.8286830-01-0.417623-0.351J42-0.135100-0.424193-0.203535

1.30000

14

-0.119302-0.1,.77130.5568510-010.4696040-01-0.6360100-020.7661800-01-0.119168-0.121094-0.5716450-01-0.360329-0.4260270.165)110.4653011.00000

15

0.476550-010.5145/J-010.5/.301/0-01

-0.2197160-010.6106J8J-01

-0.3404660-010.3060110-010.6952270-020.4/97/70-010.704L100-010.7764551) -01

-0.2221120-01-0.7993740-01-0.5023780-01

1.00000

16

0.1168070.186690.113035-0.3148650-010.5169180-01

-0.4227010-010.2353600-01-0.1120533-010.1507150-010.7051230-010.6720363-01.-3.4081c57 -01- 0.6560473 -01-0.5768320-010.8o5720*1.00000

CD

Cb

CDto

VAstAOLE

hir5fR

17

19

20

2122

23

24

2

'.1.

+.5,6810-02

).6751166-01

0.3961470-/1

e 134 167

0.67943-01

1.139171

:4=00-01

ti,3-01

0.5/091;00-0

-0.2117150-01

-0.145066

-0.267101D-01

6.3525370-01

-0.3349190-01

3p.1 7

"19

1.45

'10.

1%-'n

0.67005)0-m

0.1(.140A

-0.121509

-0.122169

0.786010

-0.812353D-01

4-V

.3&

6-q1

"-nl

-0.1

9.4.

11'1

q-.1

1-0.7s11929-01

0.7122750-17

-0.0559/90-07

-0.0015450-01

0.4636000 -01

0.9517860-02

-0.,9941q0-01

-0.e051749-01

0.415114) -01

0.93)015

-0.42v4300-01

-0.4378160-01

-(-.4.17xo0-)1

-r!.24)074')-01

-0.47)0210-01

0.226143-01

-0.6666490-01

0.5119400-01

-0.2192230-01

7c.10-4,46

0.)1(1)94n-ol

-0.17/:131.)-02-0.660690-G2

-0.1010563-01

-0.291741D-01

-0.7142593-0/

ean;4450-l1

c.1971 1i0-01

0.217'.4.11' -11

').1739)..0-01

0.21

01:,7

0-4.

11-0.74(zeg1r.)-01

-L.1143,19

-0.104144

-3.4!/1010-01

0.1354143-01

0.4.145100-01

0.263302

0.9.'611420-01

-0.46;146

-0.340461

10

0.95..,460-01

1.2-.??0-01

0.454/3'30-01

-0.2219140-01

0.Z90430-01

0.304/G217-J1

-0.22's300

-0.160553

11

1.411314n-01

0..1113690-31

-0.1307410-01

-0.2556740-01

0.5591660-01

-0.116113/

-0.647158

12

-1.6,o6m-11

-0.9640010-01

-0.3369450-01

-0.435.3440 -01

-0.5657100-01

0.1310?..4

13

-0.110150

-0.4:;536:0-01

-0.100760)-01

0.190580

..0.2034/6J-01

-0.119667

1..141)M:

14

-0.107001

-1.5540510 -01

-0.7414450-01

-0.4576970-01

0.726645

-0.1352906-01

-0.6602910-02

0.540611

is

.455.740

1.6349100-01

0.5086140-01

-0.152605

-0.1)7660

0.801492U-01

0.232426

-0.5203140-01

U.

C.587782

0.941570-01

0.116470

0.205992

-0.229741

0.8161170-01

0.338460

-0.6826110-01

17

1:00010

q.tosR10

0.118296

0.278992

-0.156088

0.8174800-03

0.515916

-0.116014

IR

lonno

0.932854

-0.5051423-01

-0.215710

-0.14228e0-01

0.8293540-01

-0.8176600-01

19

1.00000

0.6339270-01

-0.238449

-0.2077/90-01

0.9295133-01

-0.9180560-01

20

1.00000

-0.106229

-0.6240010 -01

0.236155

-0.2206770-01

21

1.00000

0.2619270-02

-0.155997

0.204750

??

1.00000

-0.5302620-01

-0.2207780-01

23

1.00000

-0.113753

241.00000

V3v1481E

25

26

27

Nu1RER

-:1.

6544

050

-11

1.4,

,,,1u

n0.418757

2-r.16114,4

).36s49

0.1)1141

1-c.25c-941n-c1

-).

25)^

17-C.2701/9

4r.3.18.)730-11

-1.141290

-0.149193

-P. 17554c0-01

-1.3355410-01

-0.416i100-01

r.1%2:440-91

-I.111167

-0.163982

7_0.13/.121

1.221114

3.21

6699

-r.)

.., 1

:74

-1.1

/191

99-0.210977

9 11 11

-r.%.,,,q4

-r.t..1..%4

-0.41ce..4

1.41047140 -01

1...2v.10

0.44.r.17

0.1::2020-01

(7..)471)7

0.954700

12

1-.1-1-440 -11

-).6,.15S1

-0.014577

13

c.)51. 70

-1.1.1449

-0.3

VI.1

,, 1

t:

C.'.05':4-7

-1.4114.'3

-0.11967S

lc

14 17

-r.9.47711n-11

..c.,,,1:261-nt

-1-.14A-457

1.ri1 -Qc.)-01

9.7^19559-)1

1.1 17449

0.m141341) -ot

4.477570-01

0.125115

18

-r.111,,4

0.7)167,1-11

0.5964910-01

10

..".110.31,t,

0.73.78q0-11

0.640931-01

-n.)14411,-11

-9.165sJs0 -9:

-;'.'441;.40-17

'1 2?

0.194(1)

--0.1?.170n0-11

-).-1/2:11,:...0-92

9.4:11410 -01

-n. 3.1.8 ^0-112

0.4717111-01

23

-C.79,613 3n-11

-0. Iscr;S2

-0.174707

74

C.964)&9

-1.163465

-0.361020

?s

1.00000

-0.545856

-0.530493

76

1.00e01)

0.985853

27

1.03n00

0 0 7

393

All Occupation Strata Black Males

VARIABLE. MEAN ... - STANDARD DEVIATION

SCHOOL 1 8.512061404 3.414289615

WAGE 2 2.366831140 0.9085147540

UN MEM 3 0.4539473684 0.4981478202

MIG<50 4 0.4265350877 0.4948448302

S.AT16 5 0.7182017544 0.4501223774EXP 6 29.92872807 12.63184550

TRAIN 7 0.1304824561 0.3370182869

PHY DM 8 2.773026316 0.8169556398

NEG WT 9 0.6918859649 0.9650420761

GED 10 2.763157895 0.7470116868

SVP 11 3.703728070 1.697336754

%13MCLC 12 0.1239243421 0.1310786869

%WFOCC 13 0.1514462719 0.1745036535

%8FOCC 14 0.22194C7895D-01 0.40550880730-01

DA/PW 15 22.68336428 40.60823057

NYL/PW 16 2.870881390 3.370130879

AVEMPF 17 0.3652142325 0.2873401759

FDSL 18 0.41644143720-01 0.92480274620-01

FOSLEX 19 0.66097931060-01 0.1016559490

ATPFAS 20 0.40800438600-01 0.18882315060-01

MININD 21 0.3101666323 0.1631280815

EDS16 22 5.646929825 4.563049152

UNXMPF 23 0.2229174342 0.3135152414

%FMOCC 24 0.1736403509 0.1997085858

M1NOCC 25 0.3025646930 0.1977858175

GDfSV 26 6.466885965 2.394528147

GDXSV 27 11.38012061 8.195603296

411

CORaELATICh MATRIX

V4R14eLE 1

N0M8ER

I 1.00000234567

2

0.2898511.00000

3

0.6325260-010.4313941.00.00

4

-0.113622-0.226186-0.131745

1.00040

5

-0.303842-0.299593-0.192565-0.8565320-01

1.00000

6

-0.579347-0.1005840.8,110580-02

-0.1989210-010.1704081.00000

7

0.266210.195230.19486 3-01

-0.10371-0.14085- 0.25308

1.0000

8

-0.177485-0.8258170-010.5325820-020.63246;C CI0.1303510.1321.82-0.5577650-01

$0/

t"

3VARI:.5LEisa.m,Eft

9 13 11 12 13 14 151.00000

16

N 1

2-3.9664930-01-c.I;(1_4..)-:;1

0.2165980.113558

0.1921720.2.;0.:51

-0.225329-0.21805 g:2=-01

0.263',660-01-0.124694

-0.613701 -010.762397 -01

0.5631090-010.200415

3 C.2411063-01 -0.157055 -0.9664550-01 -0.4148190-01 0.2434900-01 -6.2849130-01 -0.384756 -01 0.213670

4 0.1116290 -01 -0.5365440-01 -0.5651720-01 0.7670293-01 -0.9370050-01 -0.2o23000-01 0.200012 -01 -0.1563610-01

5 0.2227473-01 -0.138967 -0.116638 0.111489 -0.4265550-01 0.2903550-01 0.592703 -01 -0.6706573-01

6 0.7959900-01 -0.151812 -0.145132 0.150112 -0.6483409-01 0.2527873-02 0.751419 -01 0.1133090-01

7 -0.3487.140-01 0.120491 0.123875 -0.136442 0.578100-01 -0.1093130-01 -0.537744 -01 0.3246000-01

8 0.592040 -:N.34233d -0.187319 0.413955 -0.,66432 -0.193799 0.675164 -01 -0.1617340-02

9 1.00000 -0.193181 0.2804000-01 0.110629 -0.244834 0.104109 0.575852 -01 -0.2803590-01

10 1.00000 0.904934 - 0.501733 0.2005430-01 -0.140.399 -0.654/01 -01 -0.5048520-01

IL 1.00000 -0.483609 -0.11%937 -0.144145 - 0.59s973 -01 -0.5429370-01

12 1.00000 -0.374331 -0.7725220-01 -0.606504 -03 -0.9937010-01

13 1.00000 0.550263 -0.715793 -01 0.1694060-01

14 1.00000 -0.100166 -0.128079

15 1.00000 0.644337

16 1.00000

vs.:At:.1 17 16 19 W...49tK

1 1.15512,0-Ji a.J.3,e)-01 0.7aw56?,..3l 2 c;.,-3;sJ) ,/ '" 2: C.:36 3 J ..5 P.-.:.)-0 1 ^.. 34 14c.11) Ui 4 -0. 1J3, 1J-01 -J.6C1.-.'1-01 -u.,.432910-01 5 -3.1,,,4:6 -n525,,,%:0 -,1 -;.6...t.,-,1'.3-Z.1

6 C.:0.1940-01 0.6.-8"10-02 -c.172!,:10 -01 L.;J1,3...-01 71.0:,.,,,)-;-:( 2.114617

8 C.4263-160-01 0.75,,i30-01 0.49910-01 9 3.461'26:.0 J1 C.56 :6'50 -01 .1.'1'.451C3-01

-0.8819940-01 0.1-170-0a n.q23770-31 11 -6.255721v-01 0.145945 0.161354 12 -C.7642500-J1 -0.105.'25 -3.1175,,2 23 -0.3667210-01 -3.610:1%0-01 -0.7765830-01 14 -:.165394 -3.822673J-01 -0.107541 15 2.173551 -0.25794,1.,-01 -0.45;1.150-01 16 0.681092 0.29525)1D-01 C.6,11733-01 17 1.00000 0.135232 0.203351 18 -.00033 0.,;76992

19 1.00000 j 20

21

CA) 23 22

24

VASIA6LE hum at 6

1

2

25

-C.10e123J-01 -0.145744

26

9.236)10 0.1 :91.35

27

0.201059 0.214:369

3 -C.1195170-01 -0.1177E3 -C.126191 4 -0.3122040-01 -0.6391E00-C1 -0.6t42650-01

5 C.422,):43-01 - .3.133119 -0.131832

6 0.3111440-01 -3.152419 -0.141551 7 -0.4176540-01 0.1.'6052 0.133292 8 -C.265148 -0.239577 -0.199117 9 -3.121175 -0.4137700-01 0.2516790-02 13 -C.347:".8 0.953419 0.v41797

11 -C.451±46 0.991149 0.92q151

12 C.316625 -9.491365 -C.46)979 13 0.747023 -0./523710-01 - 0.9531820 -01

14 0.839516 -0.152371 -0.155226

15 -c.a,csliu-01 -0.457,v0J-01

16 -0.7521620-01 -0.56/3C9D-01 -0.46:5.2D-01

17 -0.116748 -0.3951560-31 -0.2772060-01

19 -C.160115 0.1254C') 0.123992

19 -C.181752 3.143193 0.139971

20 -C.7106570-01 -c.'.5393S40-01 -0.4855070-01

21 0.150163 0.2286950 -01 C.1101010-01

22 0.3422390-01 -0.1571610-01 -C.1:96!,40-01

23 -0.6147013-01 -0.9111940-01 -0.907s473 -01

24 C.762555 -0.9666040-01 -0.114607

25 1.00000 -0.428536 -0.421426

26 1.00000 0.983491

27 1.00000

20 21 22 23 24

0.14Z:82 0.637:,:73-01 0.296829 0.53/04 I-01 0.137166 G. 35c,',46 -0.25-01v -0.112,.12 0.46325 -C.7263520-01

3.432' 13 -a:34264 -C.122577 0.76J25 0.154.01 -0.3,::354d-01 0.:.457..10-02 -0 151693 -0.8;a :le )-ca -0.872,6,0-j1

-0.171,.'59 -0.31128t0-31 0.775606 -0.15z:6 -0.3117550-01

-0. l'e1061)-01. -0.4.))4t:1,-GL -0.111::,7e 0.52825 J-01 -0.7361630-01

0.1):522 -0.1015350-01 -0.04135'10-02 0.5',264 3-01 C.451v030-01 -0...5216:3 -01 -0.150560 0.50',16-.-01 0.558,/6 3-01 -0.534294 -1-71/4160 -01 -0.10c729 -0.1151970-01 0.51680 3-01 -0.192:,19

-0.52,033c-o1 0.46.'0230-01 -0.144i56-01 -0. 2:.24 -C.1514/00-01 -0.....:5112..-01 C.1166523-01 -0.15:0270-01 -0.42984 3-01 -0.1e,(20

-0.123714 0.3937620-01 -0.1960663-01 -C.57492 3-01 -0.342773 0.437:290-01 0.231177 0.4012.:3-C1 -0.62470 0-02 C.961.522

-0.13:039 0.426120 0.:7:06(.,b-31 -0.96160 J-01 0.683865 -0.392:330-01 -0.220149 0.2114010-:11 0.10002 -0.6268410-01

0.540753 -0.120109 -0.3120160-01 0.4/v/2 -0.9456120-02 0.;4Le96 -0.321539 -0.5796820-01 0.73644 -0.6560183-01

- 0.131368 -0.233201 -0.1761540-01 0.66933 0-01 -0.6523630-C1 -0.2663653-01 -0.263955 -0.4035420-01 0.13165 -0.6969330-01

1.00000 -0.259840 -0.8242250-01 0.51801 0.1.:63473-C1 1.00000 0.1962.100-01 -0.29359 0.230544

1.00000 -0.10930 0.4076020-01 1.0000 -0.2512420-01

1.00000

396

All Occupation Strata White Females

VARIABLE

SCHOOL 1

4AGE 2

UN MEh 3

MLG <50 4

S.AT16 5

EXP 6

TRAIN 7

MEAN

10.574034332.051384120

0.26716738200.47961373390.283261802629.59314764

0.1083690987

STANDARD DEVIATION

2.3160328570.98395202940.44271804980.49)85246040.450824384110.58284685

0.3110128995

PH? DH 8 1.763948498 0.6966308416

NEG WT 9 0.1652360515 0.5228954011

GED 10 3.308369099 0.7754107196

SVP 11 4.160836910 1.366598869

'AB90CC 12 0.2644849785D-01 0.3409739609D-01

1401FCCC 13 0.5430901288 0.2622150920

ABFOCC 14 0.2999356223D-01 0.4426762)860-01

EA/PW 15 15.34410537 35.00407986

NYL/P4 16 2.2232516/4 3.017452709

AVEMPF 17 0-3503766452 0.2717903175

POST. 13 C.2486631804D-01 0.3975619450D-01

FDSLEX 19 0.:3324286530-01 0.60343967420-01

ATPFNE 20 0.40006974250-01 0.1697161378D-01

MININD 21 0.4115755591 0.1995394110

EDti16 22 2.t.,61311159 4.906322707

UNXMPF 23 0.1069819027 0.2395824421

AiMOCC 24 0.5730636910 0.2682517203

MINOCC 25 0.6065321888 0.2512549256

GD+SV 26 7.469206009 2.089204571

GDXSV 27 14.68225322 7.841667910

414

C:1i,F:A7108 its.Talz

VA3IABLE

1

34

5

0T

8

1

1.03000

2

0.1823341.30000

3

-0.1905870.915420-011.30000

4

-0.2096260-C1-0.3352450-01-0.4086750-01

1.00000

5

-0.2627880-01-0.130686-0.142787-0.363114D-01

1.00000

6

-0.416514-0.1497043.7366960 -010.1173100-01

-0.4899890-011.00000

7

C.1894120.112170

-0.7008250-0-0.6S23000-010.25913.3-0;

- 0.1621671.00000

8

-0.331916-0.2379790.201164

-0.1383063-010.1055903-020.141753-0.189121

1.00000

URIAJLEkiln:71.3

9 10 11 12 13 14 15 16

1 -0.172.21 0.416291 0.305169 -0.283748 0.163575 -0.22'4983 0.116575 0.9048750-01

2 -0.121725 0.110082 0.121708 -0.127055 0.124374 -0.14/1c8 0.174371 0.279424

.3 0.1325203-01 -C.259012 -0.25447, 0.144464 -0.6258890-01 0.643301D-02 0.3551343- 1 0.8012133 -11

4 0.2112393-31 -0.2741160-01 -0.5379073-01 0.1662390-01 0.3,,67343-01 0.5142533-01 0.3510920- 1 0.5425710-31

5 0.4726500-01 -C.5740733-01 -0.04.5243-01 0.6134753-01 -0.2774263-01 0.654211D-02 0.902V.4a- 1 -0.442(.923-)2

6 0.2.20870-01 -0.134413 -0-S36116D-01 0.20092CD-01 -0.7916630-01 0.9266333 -01 -0.6908723- 1 -0.6911520-0'

7 -0.5028100-01 0.164147 0.149107 -0.130789 0.120380 -0.7410110-01 0.9405180- 1 0.7494063-01

0 0.502175 -0.533465 -0.316718 0.584633 -0.455619 0.201594 -0.190504 -0.127754

9 1.00000 -0.232299 0.266'360-01 0.457181 -0.269653 0.447154 -0.681.293- 1 -0.22144.20-01

10 1.00000 0.652250 - 3.630927 0.194563 -0.347398 0.6583360- 1 -0.2139363-01

11 1.00000 -0.425060 0.8737770-01 -0.255182 0.6751220- 3 -0.4768113-01

12 1.00000 -0.602695 0.259836 -0.7557313- 1 0.1140720-02

131.00000 0.5342530-01 0.10b733 0.2200210-01

141.00000 0.69567i3- 1 -0.119156

151.00000 0.763602

161.00000

vsalx0Lz

17

18

19

29

21

22

23

24

1D. 1z6.21

C...1',5t,,90-01

-0.2c.1-1::.-02

3.4'1,993-01

-0.1398)7

0.121,10

-0.300633-0;

0.12)411

23..n0.:73

0. 1.; ::+33

0.12It,/

9, I.ilv II

-0. 235',20

- 0. /71679J-01

0.170212

0.7729233-01

3. MI 145

0.6 :7 /4C.,- C1

0.13'7745

0. 17..;2 71

:1.7).

)1.2-02

-0. 1 )13 3,

0.119941

-0.6318495-01

40.13445)0-02

-0.4..,9994.;-01

-0.5219130-01

0.221151 -01

0.1,16:):..-01

-0.q/1315D-31

-3.717-33 -02

0.4'),)9930 -01

3.11-4713:.-01

-0.546_1:1)-C-1

-0.9.:16170-01

-3.e.',075D-01

0.3:,/:217) -01

0.902Iil

-0.2.10.67D-0'

-0.25)422D-01

6- 3.134221

-.0.3.9otC::-C1

-0.50:',C..-01

-0.c.:34%53-01

0.131139

-0.1!:215

-0.1;e314.,-01

-0.5:10(230-01

7G.e38.1,11 -01

3.1,7 t4 ::, -!:1

0.22730-02

-0.45.1140-01

-0.115111,

0.33:G040-01

0.'.o76343-01

0.105436

-3.12..364

0.':.:.4::.-:.i1

3.212742

0.t17...:43-01

0.1:,):29

-0.494'3d03-01

:).7%c974a-0'

-0..11753

.4

-J.'19',4D

0.51110,.$3-02

v.1712.60-01

-3.4)23403 -01

0.7324583-01

0.11G3J33-2,2

-C.1217172,0'

-0.1,J1..09

10

0.66221.93-01

-0.1.1475

-0.217312

-0.11G991

-0.4057180-01

.0.001N:11.7.-02-0.)6d0,1

C.132dA

11

-0.1305483-01

-3.419452:J-A

-0.127123

-0.105220

-0.3453200-01

-0.4253231-01

-0.170647

'

0.3324860-01

12

-3.57J-133D-31

0.5,82363 -01

0.190639

0.118203

-0.2252100-01

0.59.!31D -02

0.62070-01

.0.544234

13

-0.4131663-01

-0.3619560 -01

-0.147931

-0.232816

0.3422220-.01

0.3323510-02

-0.6500633-0'

0.506.117

14

-0.204047

0.34o09 :0-02

--3.5334747-01

-0.179303

0.172097

-0.346204D-01

-0.651:3923 -01

0.217320

15

0.517723

C.6669150-02-0.232o9D-01

0.7415663-01

-0.242261

0.6452392,01

0.226243

0.9285040-01

16

0.614972

0.714+9310 -G1

0.146962

0.436251

-0.311494

-0.1243V:0-02

0.312075

C.1634543-02

17

1.40000

0.2005673-01

0.I31111

0.359369

-0.244637

0.4723410-01

0.460363

-0.6256173-%,1

Id

1.00000

0.045453

0.122752

-0.206834

-0.460380-0:

0.613%260-0:

-0.33979,0-01

19

1.00060

0.354038

-0.335435

-0.8663030-01

0.123064

-0.152963

20

1.00000

-0.443967

-0.4094900-01

0.262425

-3.2.7854

21

1.00000

0. 158614D -01

-0.178:131

0.51864.9D-01

22

1.00000

-0.1269700-0

-0.246833D-02

23

1.00000

-0.9390553-01

24

1.03000

VIRIASLE

bj:0.1Ld

1 2 3 S

25

0.932517.1

0.t1057150-01

-0.4451750-01

0.5/3375J-01

-0.192.3733-01

26

0.301C;11

0.132770

-0.2G0142

-0.6462930-01

-0.709ooD-01

-0.1139o4

27

0.3E4036

0.142649

-0.276877

-0.6700640-01

-0.7t4359r.-01

-0.131..16

0.9431)2D-01

0.113422

0.154605

-0.3t0255

-0.433573

-0.3.399G3

9-0.14.313

-0.6845553-01

-0.5211150-01

10

0.5511303-01

0.93763/

0.12)2:38

-3.1-!4)630-91

(.9-13970

0.)72c11

12

-0.447415

-0.5.9920

-0.53971)

13

0.971247

C.11)2rE

0.b919:20-01

14

0..b72e4

- 0.2940 7d

-0.403970

15

0.081120 -01

0.2-37130-01

0.2242313 -01

16

0.2130 4 13 -02-0.4356180-01

-0.2991.3D-01

17

-0.9031)6D-31

0.2:1021D-02

0.2335603-02

lo

-0.2d97423-01

-0.6434520-01

-0.61213')7 -01

19

-0.14263

-0.165145

-0.153464

20

-0.227415

-0.111256

-0.9169913-01

21

0.429)350-01

-0.3652850-01

-0.9135200-01

22

-0.102632D-02

-3.2631133 -01

-0.2383463 -01

23

-0.4995473-01

- 0.175782

- 0.171670

24

0.593518

0.7o04563-01

0.4071177 -01

25

1.03000

0.8695410-02

-0.2571530 -01

26

1.00000

0.990209

27

1.00030

O C 0.1 rt 0 Cr)

-3

CJ rP 0 0 7

ARLABLE

399

All Occupation Strata Black Females

MEAN STANDARD DEVEATION

SCIICOLWAGEUN i':E311G<50S. AT16EXPTRAIN

Y

NEG hrGEDSVP

1

2.3456789

1011

9.6700251891,602972292

0.3221;1813600.142569269520.672544080627. )319b992

0.13:102015112.156 17 1285

0. 70277 078092.7262115873.470277078

2.9248902680. 6u582605720.46799 183760. 49`,07 150410.460)1772956

11. 76(.,565410. 4 321. -52930.6101,'.j22450..3:1111.."176645O. 7 1 3', 12068

i. 3.!4(.0 139812 0. S2 36164967D-01 0.417 )4J.33 76D-0 1

;;WIGCC 13 0.4609204710 0.2110161055%13POCC 14 0.127022670 0 0. 1'0,001540CA/Ece: 15 9.718452352 d.i670229NYL/PN 16 1.2617 56694 2.,,./07406A V 1:11 17 0. i30.:44:0655 0. 24/02::3:i82FDSL 181 0,2576..4 ).706D--0 1 0. -..h,1:1325D-0 11: DS LEX 19 0. 4625'170t; 120-01 19J 020-0 1

ATP 1.5 2.0 0.3273 1'3149D-01 C. 1t',0 94:lb D-01:111;1IND 21 0. 544404u008 0. 1/2 /c00116E0516 22 6.272C4J3:)2 5..).13.156382

N ;11) 23 0.0613054660D-01 0.2 1572 12631OCN: 24 0.5879521411 0. i')/.;408:388

el N 0 CC 25 0.6438367909 0.24.!:i419056t:D+SV 26 6.1061468665 1. 111.0.13046

GDXSV 27 10.27624685 6.5 )7;128941

417

4,LU)

C0000t

t8ZLZS'0

L9/2,17'9-

000001

tO-COLta.b0-

C0009

9tSt

hl

Z0-CE6CSEtO-

te-CC90tte-

tO-c7C9FLt0

.S7tt43-

10-C99St760-

t0-ClIZELFC.9-

cone't

tORSttl.0-

00000'1

Etzt

t0-(76t9t.07'0

t0-GLE9£070

Si::96C0-

70-09Prolt'0-

107t.6t0-

00000't

tt

401211'0

VrEtit'0

SfoCPt0-

C7S.:,9t'0

siczt:*e-

SOSSsVe

000CCt

01

SC7197.0-

09660-

,7r90i,0

6ACPPt.0-

607-LSt0

OIL6Ct0

176L610-

00000%

69.7.:Ct'0-

OcCett*O-

10-CfrSSL^0

'z',0:(q*O-

10-:IS*0

10-CC1e596'0

ttbt-CZ'0-

e6L(LtC

PL6,91.*0

Ch99StC

t71-r9t9-

70S07tc

7.3T,it'il-

10-CVq:.5 r,()

67160

Lf9StIt0-

L909sc0-

tOcet:c,OftC-

G6E7c7C

loUtt*c-

ZSCLLt*o

t0-CSIStf9v-

Q9c0St'0-

7tec6tc

9

C9P-C7-0-

659551.9-

ISS4qt0

10-UtteciO-te-GE106S0

L0 -crs,77t.8.0-

0t7L71)-

Z7,IL0

S

to-acterso

,

LG;csetn4vo

t0-cPvmz*c-

te-cvlogzs*o

to-apqrztzc-

t0-arbcz47c

10-(16,92790

L0-Ctti.toS*0-

LO-CL/ n(P;'0

10-caeLs2*0

10-c*:"5L.0-

t0-cpcv;,:c0-

t9-cOtet7.0-

9111,71 0-

S.:95,,1*0-

01)7p10-

SttStt.0

[0:,7tr0

66.14'0-

to-C9P9$:c5'fi-

5.476Lt.0-

;57101'0

/c1^:CUO

16:SIT'ts-

7

0617C

10-Q6SLbS.0

11;.9qs-0-

10-C1.7tCt'S'0

,99tPt.0-

791SW0

C'g:£97'0

LCS-Plt'J-

1

0p.

91

09000'1

St

tt

OSE7W0-

000001

999C0

L9P09U0-

0000rt

t)

PSL*7:'0

10-0*6,(91-0-

to-cr057re*0-

t0-C7SS7.6S*0-

to-C60[S60*0

tc-c6rv.ot*0

to-Gc9(.669'0-

tO-Ctah9S.0

10-06Y;991.0

tt.t.7c7*0-

ZECtet0-

LSOtr'e,

t9SP^Z'0

CT,SSFV0-

7t(t9S*0-

99

ft

00000 1

Lr6SviO-

SSLC41.0-

LLL.97.*0-

tS9but*O-

S

tt

00000%

Lt0/'0.*0-

00000%

cotp71.0-

ttcoor.o

V671t0- 10-SEL9Z0S0

OL

30e9ct

Pit:01'10

97/..31:7*;

L0S7

OC

OO

I

".::31:r4

11,:itttrilt:v;;

C 0 U

:3 U 0

00000'! ESt Ot't '0- 00000'1.

Z0-CCt t ZWO tO-araCt Pc16'0- 1'PLErr *0 bl9tRi *0-

i! L.17;"0- Z70ZsE "0

000707."0- 005991 '0

Z0-0r6(_,5 ", *0- I0-C(.?SgOn '0

Ct 9t7. "0- 01771 S*0

l.W,t '0- 90LEEE '0

L6L .0 96!, t Ft '0 ez 91 / ot,PtIP1 *0_

FS(.79C0 t0-0£ trIT;7 *0-

/ISS9th*C- LEI-Clot S49 '0-

Ll/q cE*0- 1.0-C.+10',011'0-

7LUPt '0- 70-c.;,t(079'0-

Z6Sf 't '0 tOt 9E',' *0-

CSF CC! '0- L66771 '0-

ZO-CEPP1 clf'LS1. '0

L0-rlEf 7:61. '0 t 0-Uf 17 t Lq '0-

E0-Gt P6S'E '0 170001 '0-

tO-C!.9! 0,Z '0 t 0-et ELrY, '0-

oZ t '0- qcoq,-9'0 Sa1707.0- SoCe;1 *0

t .0- i;1 *0

172 rZ

00000.1 Ilt1:f6G0 00000-1

Et 6E67'0- it( 0G300't 01711-2: 0- PP.O*7'0- LOL "(

t0-0Li1101q.0- t0-,11%r S0t.6St C- 10-0/.1*):1E-0- tC-.CPP7qCf'0- EO-CZ-Pf7V1.

ZO-CtLo,7r 0- *0- 767:SF 'f 10-C3n1::C0 10-0'L S16'0 S,.-

70-0":"EZEt *- t0-rf.1-0,,!,1 Ot 77 *0 t 0-0q1'4,t h0- 10-01,/0,:ti 0- LC-V.-",Sit 0- t 0-(17'tt`..t.'0 to-PA-1f Jr10 07t *C.-

lCI-Gt!i(Lf 0 te,-C/ r P919-0 rf.tiPt '0- to-Ott t:0190 10 -1lotrUS *0 L 0-Cif 0(.01 0-

0'19t.PE "0- I07rt,f10- to- c.)0St CZ*0 10 '0 *0

-.1r712.0_ SLS)17 0- 9-E,Q- '0- <417r,CP

f C trc-LF,NC Flt 7 Z1- t '7'0 -rt cz.vf- Sot LF7

7C-r71:.:660" /0-1-911 icr"G- t t - L0-pc,p( ci)L *0 t0 -fr.q..,tq -0 tO-C.LIrttr "0-

140-a,;et,"tc0 '0- f,'"?( ;C: *0- to-(:ct "t t "r`

LC -"M "1:-.160- 0;1 0- Sc,r,COt' C

Cf f-t" nC+ ti-11:7t,7:;00 10 ror P ti'iSt *0- I t n- zrzfit 0- r CC,1%71 0 1rott71 -0 t^ 172"f

76(10L1'0 t5r,',6t'0 hn cf7'0- t

L.? oz

FZ

t C7. f

cl

9: St of Fl 71 tL 1

L

L7 97 57

00E00'1. Z0-06Zt7Lh'0 CO0C0 "I.

to-CEIS(stV0- f 1001S.0- 00000'1.

Z0-0SL L9C6 '0- ZPLSfE*0- SL9tOr*C 00400'1

to-CP67Zth '0 1Z IS:7"0- 10-01,00LbZ.0 Of t SS0 "0 ' OC.0' I

SLOCt '0- 77cPt17'0- 0'7E105'0 OtbLSZ ' 0 I0 -C' T.F h' 0 ECDCC't

'r6S'St '0- $,St'i 61'0- FL It CC C OEW-5t "0 10 -Ck 4',V.,10 6cw'Orq '0

6t7SE1 ^0- 0900E7'0- t0-01ELLP1 '0 10-C1i ^0tS" 0 t0-01976.11-0 e.,.,,t:t,'0

IC-C[SfUZ '0 6t E717. '0 67CF,'S '0- 7",£17'0- L0-r0ects-c- 0c,.rS7";--

t0-e-q 0 n -0- L'f.777 *0 7S(ELt '0- t0-Gloo99L '0- l0-,:f i'f ;71 -0 t 0-Cc.770LS '0-

t 0,-(1( Ct St '0- 10-06'6U- l '0- :,,tott 'E IJ-076P5( ""0- t0-'7.,O. 0111'4' r- #-17'.t, t *C.-

i 0-1',S t t7E '0- Z 0-00 ; S6P '0 10-C° t,fl' :n-tict )1C t '0 t^ -r-,, ''tt.'.1- LC-0( c:.nt 'C

v)-r(A;mzo- to-ozpvq h'e lcZr ^t '0 10-a6q1u0 -0 t0 ....,...;:- I -0- t 1 i 7,;:

10-0£!=('7( '0 - E7'f 7'0 Sq°,,A 'E- IS It .+70- :" ( ;7: .0- .-,,t ,f .r-

to-.-r , r.,-,*o te-otozcece- to-lt",$)N 'F- te--(..,,--,zi.-. I.--,/.1:!t. -0- Lit "c' I z-

to-nGetoco Vt. 611 e- z I t.'-tt *C. 6atft '0 t^-eif -1"/.'0 ( i7nt 7'0

f PCcSt *0- t0-C97:"..1, 0( q.t. 4..Y.St 1 'e- 10-::t1,1 %,-: '0.- (4/ '9: '0-

nctt :P*o t0-(39!,/tcce 7gE 1 1 Z "4-- t0-6nStc60- to -:ctILI t *0- t( ct;,t *0-

OtC1ZE '0- 70-E7(0059*C- t0-Cq9elbi '0 l 0-06t 741., '0- 1 0-lt ",e '0- IC-E- I t f 0E 'C

66f; Cl 0- Lt 6,f t '0- lq St t- 'C E;;7601'0 1.)-716:9t9'0 L6if*c:*C.

10-01f-75'1'0- SiflIt'o- if t..,t.= '0 1.t t'r.c,!' a 7tqfts.") I tt 6.,,f 'E

St ^q97 '0 rt Z'c',Fl *0- F j...-.7'E crZ5(,t '0 13- Vt t 6C'y 'C is4 ".:7*C

7.7 LZ CZ t,t

livVIPYA

EF

tZ Oi

6: Pt Lt 4, St wt Ft 7t tt rt

t

S

402

All Occupation Strata Cross Race-Sex

VARIABLE mEAN STANDARD DEVIATION

SCUOOL 1 10.77109440 2.9100284642 2.953421053 1.517826903

UNioN 3 0.3642439432 0.4813180476MIG<50 4 0.4298245614 0.4951543409S 16 5 0.2861319967 0.4520462749Exp 6 28.37176274 11.62024710RACE 7 0.8604845447D-01 0.2804941750SEX 8 0.2878028404 . 0.4528333252TRAIN 9 0.2017543860 0.4013936397Ruv Dm 10 2.210944027 0.8625740317NEG 4T 11 0.4482038429 0.8344037001svp 12 4.853968254 1.878014896.5.03mocc 13 0.4767126149D-01 0.5756481215D-01.4wFocc 14 0.2745614035 0.2727570487

A3FoCc 15 0.190296574 8D-01 0.412437683813-01

DA/pw 1t 27.63948124 65.831007.86

AV P? 17 0.3995127541 0.2919289129Fasc 18 0.3538361421D-01 0.6429166881D-01AeppAs 19 0.4308847110D-01 0.175878961713-01

XmINIu 20 0.3423889435 0.1894500734EDs16 21 2.837510443 4.806049936

uNxNRF 22 0.1734650654 0.2933291042:4 i1GCC 4 .) 0.2935910610 0.2873789668NmINoc 24 0.3412623225 0.2748465174aAxsEA 25 0.25480367590-01 0.1576118492

420

C0RRE..LTIUM MATHix

VARLAJ!.?WJKSi.4

1

23

67

8

1

1.30000

2

0.3608b61.00000

3

-0.1028530.1702123-011.000J0

-0.8714660-01-0.1536660.595450-011.00000

5

-0.165901-0.15.053-0.120053-0.6614590-01

1.00000

-0.457394-0.1304610.8127710-010.4070460-010.297107D-011.00000

7

-0.150629-0.1605760 2156100-01

-0.1636363-020.2210010.409060D-031.00000

-0.4924353-01-0.40.412-0.140280.7240680-010.3645000-010.6963560-010.5344490-021.30003

VARLA3LEaum883

9 10 11 12 13 15 16

1 3.234175 -2.321)31 -0.1d4t,39 ).344472 -0.171431 0.4671,10 -01 -0.176372 0.71,4920-01

2 0.1v:505 -J.7120,..50-01 0.12..00-01 0.37,1.39 -0.131714 -0.251422 -0.233203 1.155126

3 -3.506600-01 0.2?6920 0.516:0;D-01 -0.156517 0.190692 -0.155247 -0.2167910-01 0.6C,..52;10-01

4 -0.116v:3 3.3124330 -01 -0.9!;030-02 -0.7964580 -01 0.5644663-01 0.4335540-01 0.2802330-01 -0.921';70-02

5 -0.6725320-01 0.6464210-01 0.753:500-01 -3.105523 0.114181 -0.5133170-02 0.6051.610-01 0.2151r.30-01

6 -0.301977 0.115579 0.8572140-01 -0.9571810-01 0.130163 -0.2197960-01 0.7516120-01 -0..1310010-01

7

a-0.530463D-01-0.144860

0.11207-0.330946

0.9225620-01-0.186701

-3.214124-0.257678

0.301917-0.169314

-0.3816160-010.667509

0.1732020.280526

-0.4577(.20-01-0.12)483

9 1.00000 -0.3469180-01 -0.1602210-01 0.232347 -0.132/66 -0.4215260-01 -0.102366 0.6473550-01

10 1.00000 0.526995 -0.9146270-01 0.534481 -0.552900 -0.5319890-01 0.1847690-01

11 1.00000 0.113308 0.281576 -0.353346 0.0716400-01 0.5947720-01

12 1.00000 _0.517724 -0.330603 -0.331953 0.7396260-01

13 1.00000 -0.342580 0.8362900-01 -0.3440790-02

14 1.00000 0.288422 -0.9038320-01

15 1.00000 -0.7755530-01

161.00003

c.

4.J0U

00000't

SZ

OL99SL*0

9Z

9LLtItt"0

CZ

LO-CSMOS.0-

ZZ

10-0SELL9S'0

1Z

LL9051'0

OZ

tC-010199L'0-

6t

to-rrtPtcco-

el

Le-0094,*L'0-

Ll

1C-OZ.St<60-

91

91111C-0

St

LnfOt"0

hi

LC -G' '?f'0

Et

tl,C:L'O-

Et

L0-OnQuo,-0

Li

7C-C,,tctlyz-0..

01

tC-Ctf,9c-t7'0-

6

Lq(**',Z*0

P

6fSZS*0

70-060S%P*0

9

LC-CZ0St6C0

S

Z0-0,,A0P190

9

to-erNszr-o-

E

69SCL'O-

Z

te-C9r.tEPS*0-

I

67917rm

SZ

319YIRTA

000001

L10096'0

00000'1

FIZ

CZ

t_

ZSCI.SL'O-

C194Lt'0-

t0-00t597.C.0

10-0990Z6U0

00000'1

tO-OLS6659.0-

000001

ZZ1Z

NI!

05,499E'0

91Z96£'0

06SEZZ'0- L0-0ZS9LS'0

00000'1

OE

toSC7.991'0-

606SL1'0-

1PL96Z'0

10-094998L'0 -

99P7,07,'0-

00000't

6t

C9LELW0-

LOtSOL'O-

10-006P169'0

t0-ei.p0611.0-

sczazvo- E0-009hL16.0-

00000'1

P1

01'3E591'0-

SfiLLWO-

S6999G.0

t0-COtZt9t'0-

67/.6E7,'0-

7,C6LEU0

10-0SCEL66'0

00000'1

Lt

SSOZOL'0-10-0071696'0-

PCL'ICZ'O

l0-OCOCSZh'S

6C59ZE.0- 0-09C9LL0.0-

10-0660CC5*0

77tES9'0

9t

p?sr,It'o

17v7LI40

10-09FEL'0- 10-VEL9C7'0

Ph9L6E'0

,t'qECL1'0-

t0-09LSIS'0-

(69%Lt'0-

St

cq

0,76C96'0

(15061'0

Ott991.0- 10-0ZIOELZ*0

OtZt:.C.0

1Et6S1.0-

6117,01'0-

P6PZZI*0-

6t

ZS6Ltt'0-

PtIELCO-

LOLOZL*0

LO-CtLPOOL'O

6f4ZIt'0- to-t7SE;:;01'0-

t?-ccf..etl'o- i0-cofbriP'o-

it

9LZ969.0-

91069(.0-

L0 -GO'Nt)IP'0- to-cocetwo-

bst,:ft'o- to-ctLtobto

10-0.66,)E6C0

70EcC1*0

Zt

0,

POL817'0-

CLSZZL'O-

oftrnh'0-

10t,'ES':-

10-C66C:16'0

10-0,At61 :'0

ri09c7,*0

10-CLhttSi'0-

L9CEIL'O- LO-OLChLOCO-

958P91'0- ZO-OOLOWO

1.1-016L0CU.,

ZO-0/.P9tet'e

10-r6E6SW2

70-0t751517'0-

it01

Z69,41*0-

folZet'0-

70-0I;;X61'0

L0-0t9LItE'0-

10-0tZcPt6'0- 10-05N'Z'O

f3-1/716t('0

S:1/.0.0

Wth9'0

99Ctit'0- 16-r)uSq9h*0

S49694'0

cw.GS!'0-

fl,rft.0-

i=:00-

:

to-ctP.,:is'o

to-ctzgrwu-

90-0kuin'0-

it.119L'0

10-0cM-E'0

LC-0SNIW--*C-

ro-;:,qh1c-vo

LO-OEILf:'P'C-

L

L0-Ort:at'0

10-019U,OL'0-

t0-GWC9E*0

10-oCZ901/'0-to-ctuotvc

10-Q146f:39'0-

t-C,(1Eft'0- tO-OStol"0-

9

t0-CN011i,Z*0

ZO-TASLOC'0

:0-(1167119'1,-

1S/ZE6'O

L0-OU070C9*0

6791.01'0-

tit-Ch40967'0- tC-0(trcf50-

5

t0-001,416,'P

10-0:09149'0

10-39170:U0

10-067nSC0-

tO-C1rOtS9'0

ZO-0CILLC6'11-

10-09SFqW0- 10-0PLSL11'0-

9

GCCItt"0-

09tOSL'0-

Et+91',1'0

!",,,,7!*0-

S1'751'0-

vztvri..(

:0-c-v.s1;1=yb0

sopwo

t

ci:ISs75-0-

7/c-Lre-

10-0/PL9C0

L0-C%G0C.C:-.0-

9 q:16:7'0-

tr,o;c:bo

aqvico

Pff.370

z

to-cwIfitso- 10-ce1061'0

tpnct'o- to-lizczwo

10-197,):90- 1C-OKCZW0

til-Ilft.C.LIE'0

CCOSSL*0

t

F741:;'w

hZ

CZ

ZZ

It

OZ

61

81

LL

7101IEVA

BMIOGRAPHY

Arrow, Kenneth. Some Models of Rac'jal Discrimination in the Labor

Markel. Santa Monica: R= Corporation, Publication RM-6253-l.C,

1971.

Ashenfelter, Orley and Mooney, J.P. "Some Evidence on the Private

Returns to Graduate i2dueation." Southern Economic Journal 35

(January 1969): 247-256.

Averitt, Robert. The Dual. Economy: The Dynamics of American Industry.

Strw-ture. York: Norton and Company, 1968.

Becker, Gary, The Economics of Discrimination. Chicago: University

of Chicago Prass, 1957.

. Human Caoital. New York: National Bureau of Economic

Research, 1.964.

and Chiswick, B.R. "Education and the Distribution of

Earnings." American Economic Review 56 (ay 1966): 358-369.

. "Hmuan Capital arid he Personal Distribution of Income: An-Analytical Approch." W:S...1v1,-:,vtini,ky Lecture No. 1. Ann Arbor:

University of Michigan, Institute of Public Administration and

Department of Economics, 1967.

3ondix, ileinhard and Lipset, Seymour M., eds., Class, Status and Power.

G.Incoe: Free Press, 1953.

1erg, Tv.a. F.ducation and Jobs: The Great Trainin:; Robber,.. New York:

Praeger, :.969.

Bjerke, Kjeld. "Incomo and Ti,agc Distrihntions--Part 1: A Survey of

tho Literature." Review of Incol-re and Wealth 16 (September 1970):

235-252.

Mang, M. "The Privata and Social Returns to Investment in Education:

Somi- R(.!sults for Great Pil.itain." Journal of Human Resources 2

(Spring 1967): 330-346.

Blinder, Alan S. "Was...e Discrimination: Reduced Form and Structural

Estimates." Journal of Human Resources 8 (Fall 1973): 436-455.

405

423

406

Bluestone, Barry. "Low-Wage Industries and the Working Poor." Poverty

and Hunan Resources Abstracts 3 (March 1968) : 1-14.

."Economic Crises and the Law of Uneven Development." Politics

and Society 2 (fall 1972): pp. 65-82.

; Murphy, William; and Stevenson, Mary. Low Wiees and the

Workin,, Poor. Ann Arbor: Institute of Labor and industrial

Relations, University of Michigan-ayne State University, 1973.

Bowen, William. Wage Behavior in the Postwar Period. Pr inceton:

Princeton University, Industrial Relations Section, 1960.

Bowles, Samuel. "Ttriards Equality of Educational Opportunity?"Harvard Educational Peview 38 (Winter 1968): 89-99.

"Tounrds an ducational Production Function." In W. Lee

Hansen, ed. , Fduca,:ion Inr_or-e and Human Capital . New York:

National Bureau of Economic Research, 1970.

."Migration as Invc'stment: Empirical Tests of the Human

Investment Approach to Geographical MobiJity." Review of Economics

aad ics 52 (Novemier 1970): 356-362.

."UnecInal Education and the Reproduction of the Social Division

of Labor." Review of Radical Political Economics 3 (Fall-Winter

1971): 1-30.

. "Schooling and Inequalay fron Generation to Generation."Journal of Political. Economy 80 (May 1972): 5219-5251.

. "Understanding Unequal Economic Opportunity," Afir2ricati

Economic Review 63 (May 1973): 346-356.

Bronfenbrenner, Martin. Income Distribution Theory. Chicago: Aldine-

Atherton, 1971.

Carroll, A.B. and 7t stern, L.A. "Casts and Returns for Two Years ofPostsecond.lry rechn-cal Schooling." Journal of Political Eccnorly

75 (Decemoer 1967): 862 - -873.

Clark, John Batts. The Distribution of Wealth. New York: MacMillan,

1900.

Coteman, James S., et al. Eauality of Educational Op2ortunitv.

Washington: U.S. Government Printing Office, 1966.

424

407

Council of Economic Advisors. "The Budget elessage of the PresiJent,"

The____Budet _of the United States Government, Fiscal Year 1970.

t&whington, D.C.: U.S. Government Printing Otfice, 1969.

Cummings, Laurie D. "The Employed Poor: Their Characteristf.s and

Occupations," Mont111yLII)or Rev' cow 68 (July 1965): 828-38.

Delchantz, George and Evans, Robert, Jr. "Low-Wage Employment: An

Inventory and Asses:n Northwestern University, ,.d.

(mimeographed).

Denison, Edward F. The Sourcer_ofFcononic Growth in rhr United States

and Alter:tat:c.v.-2s PeEore Us. 1.,ew York: Committee 1:er Economic

Development., 1962.

Dooringer, Peter B. "Manower Programs for Ghetto Labor Markets."

2roceedig_ofth 21:;c_Aanual Win:er_Meeting of theIndustrial

Relli.Lea$ Rcsearh A,;seciecion 21(Decumber 1968): 257-i67.

.and Pi)re, Michael J. Internal Labor Markets and sliviower_

Analws:is. Lexington: D.C. heath. 1971.

Dunlop, John., 4:412.i Doterminition fluder Trade Unions. Nev. York: John

Wiley 1944.

. "Productivity and the Wage Structure." Jr leceme Fin)lovmer,t

ziadlublic Polio?, pp. 341. -362.. Edited by Lloyd A. Metzler. Ncw

York: Norton, 1948.

Durkheim, rmlle. On the Division of Lahor in Societi. New York:

Mecmit1,-,n, 1933,

Edgeworth, F,Y. 111.qual Pa.: to Men an Women." Economic Journal. 32

(Deccmbr 1922): 43J -457.

Farrnr, Doraij E. and Glauber, Rohert R. "Mult.ic_ollinearity in

RegrJas:on hnalysis: The Problem Revisited." Roviewof Economics

dnd St:tzi,itic:; 49 (February !967): 92-107.

Fine, Sidnc: A. "...no IRe of 'The Dictionary of (,)cupntionnl Titles'

As A Source of ;-;tima'..:s of Educationsl and Training Requirements."

- ioamon '..'coerces 3 (inter 1968) : 363375.

Priedran, Milton, Price Th_orv: A Provisional Text. Chicago: Aldine,

1967.

Galloway, Lowell. "rhe Negro and Poverty." journal u_f ;iusiness 40

(fanuary t967): 27-35._____

408

Garbarino, Joseph. "A Theory of Inter -- Industry Wage Structure

Var iat ion." (quarterly Journal of Economics 64 (May 1950) : 282-305.

Goldner, William. "Labor Market Factors and Skill Differentials in

Wage Rates." IndustriajRciations_Re-=earch Association, Proceedinvs

of the Tenth Annual Meeting 10 (December 193S).

Cordon, David. Economic Theories of Poverty and Unde-romplovment.

Lexington: D.C. Heath, 1972.

; Edwards, Richard C. ; and Reich, Michael. "Labor Market

Segmentation in American Capitalism." Cambridge: Harvard

University, March 1973 (mimeographed).

Green, Alfred L. Manpower and the Public Eme_ ent Service in Europe.

New York: New York State Department of Lab, 1966.

Greenslade, Rush V. "The Economic Effects of Collective Bargaining

in Bi...:JMillOUS Coal Mining." Ph.D. dissertation, University of

Chi,-:ago, 3952.

Ilanoch, Giora. "Personal Earnings And Investment in Schooling."

Ph.D. dissertation, University of Chicago, 1965.

."An Economic Analysis of Earnings and Schooling." Journal.

ofiinmIln Resources 2 (Summer 1967): 310-329._______

II nsen, W. Lea. "Total and Private Ratess of Return on favectment in

School ing." Journal of Political Econ.ny 71 (April 1961): 128-140.

_ :Weisbrod, Burton A. ; and Scanlon, William J. "Doterminants_ _

of 1-Innings: Does Sch.)oling r:eally Count?" M'dison: Discussion

Paper, Institute for Research on Poverty, University of Wisconsin,

August 196t (Preliminary Draft).

and Weisbrod, Burton. "The Distlibution of Costs and Direct

32a?Iits J: Pub3ic EdIcntion: The Case of California."

j:ounal _or_ t;umdr_ Rer,ources 4 (spring 1969): 176-191.__

Weisbrod, Bt:rton; 4,nd Scanlon, W.J. "Schooling and Earnings

of Achiev,rs." Amer icon Fconomic Review 60 (June 1970) : 409-418.

Harr:Fen', 110.,nett. Educdt on TrainintL2nd the Urban Ghetto.

Ddltimore: John Hop'cins Press, 1972.

Hay,,, William L. Statisticsfor_Psvcholeilists. New York: Holt,

Rinohart, and Winston, 1963.

Hicks, J.R. The Theory of WApes. London: MacMillan, 1932.

426

409

Charles C. and David, Martin H. "The Concept of Job

Vacancies in Dynamic Theory of the Labor Market." Madison:

Social Systems Res..aieh Institute, University of Wisconsin, 1965.

Routhakker, H.J. "Education and Income." Review of Economics, and

Statiscics 41 (February 1959): 24-28.

Hubbell, Robert and MacDonald, Martha. "A Study of Employment

Recruitm,?nt in Boston." Social Welfare Regional Research Institute,

Boston. Collnge, forthcoming, 1974.

Jencks, Christopher. ineoualit,- A R;,aspssment of the Effects of

Family and Schc,o;i-,g in Amccicl New York: Harper and Row, 1972.

Johnson, George E. and Stafford, Frank P. "Social Returns to

Quantity and Quality of Schooling. " Ann Ac her: Department of

Economics, University of MIchigan, 1970 (mimeographed).

and Youmans, Kenwood C. "Union Relative Wage Effects by

Age and Edica:iin," industrial and Labor Relations Review 24

(January 19/i): 171-179.

Kane. Edwird J. F.conom;c. Statistics and Econometrics. New York:

Harter and 1:uw, L968.

Katzner, Donald A. "Theory and Cost of Racial. Discrim4nation."

Ph. 1). dissertation, University of Pennsylvania, 1966.

Kerr, Clark. "fhe E-ilkanization of Labor Markets." in Labor Mohilitv

21nd......:ic,112portentty, pp. 92-110. Edited by E. Wight Bakke.

11.e New le-1-,nolo3r Pros: and John Wiley, .195/4.

Klein, L. R. An II:trod:lotion :o Ecene,petrics. Englewood Cliffs:

Preliticc-:;..1.1, 1962.

Kahn, 1l'/as . Tho Strue:or, re. Seiontific Vevolutions. Chicago:.....

Jni.icrsi:y .'f Chlvaio Press, .1962.

Leontief, "ftc. Strue..ture of the U.S Economy." Scientific

American Apr.I 1935, pp. 3-13.

Lester, Ri,.hard. "Short:om:Tvi of Marginai Analysis tor Wage-Employmcnt

Problems." .%:.:(rican 1%conomicf Roview 36 (March 1946): 63-82.

. "Margini,ltsm, Minimum Wages, and Labor Markets." American

Lconsritc Izeview 37 (March 1947):' 135-147.

. Mancw:(fr Plaanin" in n Free Societi. Princeton: Princeton

Univetsttv Presq, 1966.

4%7

410

Levinson, Harold M. Unionism,ya?,,e_Trouds., and JIICOLW Distribution

1914-1947. Ann Arbor: University of Michigan Press, 1951.

. Determining rorces in Co11c!ctive. Wage !in in. New York:

John Wiley, 1966.

"Unionism, Concentration, and Wage Changes: Toward a Unified

Theory." Indostrizd and Labor Relations Review 70 (ionuary 1967)

198-205.

Levitan, Sar A. "Manpower Programs Under Republican Managoment."Poverty and human Resources 5 (March-April 1970) : 4-13.

and Mangum, Car di L. Federal Training ind Won- Prmrams in

the Sixties. Ann ixbor: The University of Michigan, ln3titute of

Labor and lndustria. Relations, 1969.

Lewis, H. Gregg. Cniynt and Relative Wo,'es in the Unitnd States.

Chicago: UnivPrsity oi Chicago Press, 1963.

Machlup, Fritz. "Marginal Analysis and Empitical Research." Am,:rican

Econcmi(' Revie.w 36 (September 1946): 519-554.

."Maignalism, Minimal Wages, and Labor Marl;ets: a &ejoinder

to ;'it Antimarginalist." American Economic Review 37 (March 194?):

148-153.

Marglin, Sto.phan. "What Do Bo3:,es lo?" Peview of Radi.-:.a1 Political

oonomio.s 6 (c17:.1.mer ''orhceming.

Mtse:s, St.xiley H. "rhe E,tect. of :,amily Income on Children's

bducat1on: Findings oa Inequality of Opportunity." Journal

4 Spin; 19(7)1: 158-175.

Miche1s,m, Stopttan. -Income:, of Minorities. Washington, D.C.

1968 (!aimeogr;-Thed).

Mill, J.S. The of 2oliticiil Econmv with of Their

So::: 0 IiiloscnEv. En C,livcted Works

Joirn -_y,1:. Nil]. Yoranto: Univil..rs1ty of Toronto Press, 1965.

or, HermAn P. "Ahnuil .1nd Lifetime Income in Relation to Education."

Amerieln rclnemie. v ow 50 (December 1960) : 962-986.

Mincer, Jacob. "On the Job Tra;nim:: Ces's, Returns, and Some

Implications." journak_oLPolitical yconolor 70 (October 1962):

50-79.

LIZs

411

1,11 [Icor, Jacob. "The D is t_ rihu I or f 1.aitor I ncome!t: A Stirvey." 101naof Econor. ie I. i tera Lure 8 (March 1970) :, 1-26.

Mitchell, Wo:tley C. j'y2esof Jbeorv, Vol. I. NOW York:

Augutttus I. ke 1 Icy, 1967.

Morgan, James N.; David, X.:rtin H.; Cohen, Wilbur J.; and Brazt?x,Harvey E. Incone and Welfare t1,0 ted States.McGraw-ill II , 1962.

New York:

and Si rageld in, 1s;::all. "A Note on die Qealjtv N:s ion itt

Educ:.1 Lion," .1.7urna o 1.itic.jl cons.rav 76 (Sep 1 r tiOcts-ber 1(368)1069-1077.

Mortenien, Dale 1. "A Theory of Wane and Employment :)vnamics." In____Mic roeronon: Foul,t1,.!;_ ion c.f.' ;.:ttolo,,m-ct. and I'nfl a t. toe Theory,

pp. L67-211. Ed I 'oy Lsit.oint_i S. Vto1 ps . Mew York: ';or t on, 1970.

Nichols, Don.ild A. "Nat:1,ot Clear ing for IF:terogeneous Cnp4tal Goods."Mi oe.2enorti ,74 and in flat ion '11;e0u,

pp. 394-410. Ed icet: by Ed:1:1a:: irnelps. Nt_w Yer!:: ;41:.- ton, 1970.

W., Walter. "Labor as a Quasi-Fised Factor."Econo:71cs 70 (Decembr 1962): 538-553.

lournql Political

Quick, Paddy. "Women' =3 Work." R.'.' of Rac.:i cal Pol EconoNics 4(July L972) : 2-19.

Rees, Albert . "Union '.:age Ga , and Eatcrprise non000l y." 1'..sflays onml P t ions Research: Ann Arbor: : iln iversity of Mich igan,_ .

institute of i,,00r and Induri,-.1 Relations, 1961.

and Shull_ 4, George B. Wor1:e rs ond Wnn..--; in !i:: ;:rb.1,1 t al,or.-_, _""o. 1'a u: in hr-,r "it v ol Chicapo ress 1')70

Rensb,t,', "Est i Ian r L rIt i?t '.tt int s to Ecitica Ii cit." Rev i oricc 4 icon 1960) : 31.8-324.

Reynoldi, :Axyd.

Row, 19f.i.ae St rtu.t..1,-e of I..--hor ::ew York: Harper ird

-------

:lad -fa ia H. T I vel tit I, nt of Waoe rut; Lure . New

Haven; `.! ale 1.1: Lye rs t kr Press, 1956.

Rib I a, no,. as I . 1:0ticat_ton and Pover cv. Washi nnton, D.C. : Brook ingsIn =t. tu I ion ,

Po') i nun , A .G. Ind V. , E. , ed s. The Eco ro-,Nm ics at Ed tic a I. ion.1 cittl.;n: St. , 196o.

412

Rose, Gerald. "Differential. Returns to investments in Human Capital

in the Academic Labor Narket." Ph.D. dissertation, University ofCalifornia, ilerk2ley, 1969.

Ross, Arrhur M. TradeUuion Berke ty: University of

California, 1948.

and Ooldnar, William. "F6rces \ffecting the interindustry_______Structure." Ouarterly Je_,..irnal. of Economies f)L (may 1950): 254-261,

Roy, A.D. "fhe Distribution or Earnings and Individual Output."Economic Journal 60 (eptember ).950): 489-505.

Schultz, T.W. "Invest:A:Au_ in Man: An Economist's View." Scp:ial Service

RQVieW 33 (June 1939) : 109-1.17.

"(nvestment in Human Cap11.al." American Economic BLview 51

(March 1)61): 1-17.

Cayo. Education a:1_Inco.ro. New York: Viking Pra:,s,

1964.

Siegel, Paul- "On thc Cost of Being a Negro." Socioloqal locuiryWinter 1965.

Sliel:ter, Sumner. "Notes on the Structure of Wages." R,../Jew of

Pconom1c. 11d ctaristirs 32 (ebew,ry 1950): 80-91.

Sobotka, Swp..en. "11.e Influen,e of Unions on is and l7:arningn of

1,abor i the Construction industry." Ph.D. disertation, University

,.f Cuicago, 1952.

:1orokin, ritirim P. ,11,1itv.. New York: liarpor, 1927.

S affard. F.can': F. "CcnvPrtratioti and label- Earning3: A COTM:11.4

lcon:mL! :eview (March F:68): 17A-181.

)ea.e An 'ncrodoc..ion to QuIntitative E,-onomic_ _ _ _ _ . _ . _ . _ . _ . _(.1::co: Rand M,;,..11.y, 1963.

Sz,rv:,!y of_ r:k !tz:(1 i son: S. E. 0 . ilwiso, Dal a

Comptla: ion Cai. 'r Soi,il Science Buil.iing, The University

of `sc..,n.:in.

Thurw, Lestcr C. Poverty Ind Discriminat ton. Washington, D.C.

:ro.Aings Institute, 1969.

Puma!) tal Beimont: Wadsworth Publ fishing

Comr any, 19i O.

4J0

413

U.S. Congress. Joint EcoaoiiLc Committee. JlleAmorican Distrii)ntion

of lecomc: A Structural Problem, by Lester Thurow and Robert I. B.

Lucaq, JoiaL Co.o.nitreo Print. W:Ishiagton, D.C.: Government

Printing Office, 1972.

. Joint Economic Committee. Polo rat Prof for the

Develou'.:ent of Human Resources. "The Occupltional Distribution

and Returns to Education and Experience for Whites and Negroes,"

by LestLr Thurow, Joint Comittee Print. Was}-int., D.C.:

Goverament Printing Office, 1968.

. Joint Fonomic Committee. Postwar Movement of Prices and

in s2l:alufactrit-0,' inc'e,,tries by Harold M. Levinson, Joint:

Cox.,nittee Pr ott, SZacty P&per 21. Washington, D.C.: Government

Printing Office, 1960.

C. S. Department of Commerce. Brreau of the Census. United States

Census of Population: 1960. "Occupational Characteristics."

Seri PC(2) 7A, 1966.

. BuroNu af the Census. United St.-,tes Census of ropolati.-n:

1360. "industry Chatactetistics. Series PC(2) 7, 1966.

.Current. Popul ttton Renorts--Consumer Income Series P-60,

No. 69, Apr i.1 6, 1970.

.Currcnt Ponn1-(tion Reports: Consumer incom. "Characteristics

of the Low-Income Population 1970." Series P-50, No. 81, 1911.

De.)artmest Labor. ASup-plent. othe Dictionoflocerar.,eanl TitiQS. Washintor., D.C.: U.S. Government Printing

Off,c, 1966.

. Mon:ower Administration. "rho influence of gDTA on Earningu,"

.ower EA,aluaion P.cport No. 8, December 1968.

. Yanp,wer Adis:Istration. "Ir:provements eel. '3 in Cont-ca.:ted

or on-tna-jo:, um:.er ne Manpower Development and

Training Aot." 1968.

. of :rmntion. "Thr,-,e Stat-u:atds of Living for an

!Thal Vaiiv of Four." '.;pring 1967.

' 't.or rii.ace of Pam( I y Heads, by Povert v Status of

mr. t y , 195r;." M-lo2LoweLIZe.pott.. oLt.11e Pres ident. Wash lug ton:

U.S. Government Pri L Ot II(' r, 1970.

Wachtel, Dawn. The 1.:oikin;.: Poor. Mimeographed. Ann Arbor:

in,tituti ot Labor and Indu,trial Relations, University of

Michigan-Wayne State Cniversity, 1967.

431

414

Wachtel., toward and Betsey, Charles. "E.mployment at Low Wages."

lhe Roview of Economics and Statistics 54 (May 1972): 121-129.

Waldman, Elizabvth. "Educationai Attainment of Workers." Monthli

labor Review 9.). (February 1969).

Weber, Max. "Class, Status, and Party." In :lox:Weber. Ess.,:xs in

Sociolop.y. %ew York: Oxfotd University Press, 1946.

Weiss, Leonard W. 'Concentration and Labor Earnings." American

Economic itc.view 56 (:larch 1966): 96-117.

Welch, Finis. "X:,asurt!ment of the Quality of Schooling." American

Ecoomic Rcvi.2w 56 ;May 1966) : 379-392.

Wootton, Bachara. The Sociol Fouudations of Wal:s Pol icy. London:

Litwin Univcrity Books, 1955.

432